UKH

Three quarters of a year of these posts….  Didn’t expect this to keep going so long.  It’s the Duracell Bunny of plotting.

Not much interpretation this week as we’ve more or less been in a holding pattern for the last week, cases/day roughly steady at a high daily rate.  The critical question becomes “is this because we’re at the tipping point for herd immunity” or “is this because we’ve reached the steady state where people are loosing immunity and getting re-infected”.

• Given the last ONS antibody survey and estimates at R0, the former (herd immunity tipping point) seems more likely.  If this is true, cases/day will need to start dropping a lot in the next 2-3 weeks I think - certainly in older adults (say > 45), schools conflate things a bit below that.
• The data I've plotted here can't answer the question though.   The dashboard really needs to categorise data by immunisation status to maintain its usefulness.  Qualitative snippets in the news suggest hospital admissions remain in the mostly unvaccinated, and quantitative data from somewhat comparable nations (very loosely in terms of vaccine demographic uptake) is similar; last week I showed a lot from San Diego county about vaccination status of cases; this week I’ve attached a plot from them on hospitalisations, sourced from [1].  There may be more locally relevant quantitative information on this in SAGE documents, I’m afraid I’ve been out enjoying some late summer walking not reading SAGE stuff…

The 7-day change method for measuring the rate constant in England shows that we’ve bobbed back over the line from mild growth in to mild decay - although that last data point could be nudged back to growth by a bit of lagged data, but the trend toward decay remains clear.   The plot annotated with the Central England Temperature has rate constant and temperature moving in opposite directions (as we often see) for this latest wobble in rates.  Unusually its happened with both a high absolute number of cases/day and without being overlaid on a rising or falling baseline; this makes it easier to see that the rise-then-decay in cases translates through to a similar behaviour in hospitalisations - this is particularly clear in plot 18.1.  This looks to me like a solid confirmation that these weather associated deviations in the case rate over timescales of ~7-14 days are a real indicator of a real change in infections.

At Wicamoi's suggestion last week, I have updated the Scottish 7-day rate constant plot to show their latest dropping of big restrictions; it feels a day or two soon (in the data) for that to translate to the exponential rate growing again there.

The next predictable, big events are the return of schools across the nations, starting with Scotland next week I think.  

The next piece of scheduled data I’m most awaiting is the ONS antibody survey update due on August 19th; when that’s out we’ll have a clear context for if we expect cases to be continuing to however around growth/decay or if they really should be in to decay now.  What really matters now I think is “cases in people without pre-existing antibodies” and that data doesn’t seem to be accessible.  I hope its being compiled and analysed somewhere in details behind closed doors, because it’s that that’s going to flag up if our current policy is going to work out or if we're going to have to re-think things a bit...

[1] https://mobile.twitter.com/SanDiegoCounty/status/1425909612239196162

Link to previous thread: https://www.ukhillwalking.com/forums/off_belay/friday_night_covid_plotting_38-...

The English plots

The effects of the recent “weather wobble” in cases are quite subtle as, and the trendline on plot 6e is quite smoothed, so you don’t really see the weather effects here, just that cases bottomed out after the Euros turned off, and have been gently growing since - on average.  That bottoming out and rising translates visibly to admissions (7.1) and occupancy (7.2); the peaks and valleys of plot 6 are spread out by the variable time it takes for the illness to affect different people and land on plot  7.1, and then they're smoothed over again by the nature of hospital occupancy (7.2) which is a bit like a running average or box filter (admissions get added and some time later people leave).

The current prominent fall in daily deaths is probably reflecting the big drop post-Euros and presumably won’t go on much longer as this measure stabilises at a level corresponding to cases and admissions.   

The Scottish plots.

Both cases and admissions look to be rising, the data is pretty noisy however.  Definitely one to check back on in a few days as their latest unlocking fully lands in the data.

The four nations plots.

The “football cliff” looks much smaller in Wales and Northern Ireland; not unsrurpsing perhaps. 

All 4 nations are now in some sort of sync, with cases hovering somewhere around the point between decay and growth, and currently all turning back to growth.

The demographic cases plots for England

Not much change form last week where I noted that the rise came harder/earlier in the age bands 15-30.  That early growth has concentrated the relative distribution of cases back in to these bands (for now); most of the shuffling in the probability distribution is with the younger 10-15 band so I wouldn’t expect this to manifest as a visible change in hospitalisation or fatality  rates.

Two ideas over why growth came first to the 15-30 band, that data neither proves nor disproves either:

1. This growth is driving growth in older ages
2. This age band is now a more sensitive "canary" to conditions conducive to growth in cases

Option 2 feels more likely, as the data has never shown support for option 1 before, and we know this is the age range with the least vaccine uptake and the most engagement with nightclubs etc.  So, we might expect it to turn to growth sooner and exit growth later in a cyclical temperature wobble.  Self limiting after a while as infection wins antibodies... ?

The English demographic rate constant plots.

It's the heat map on 18.1 where the translation of the brief period of growth in cases (vertical gold bar across most regions on the right edge of the top left subplot) can be seen to translate to hospitalisations.

The UTLA plot gives hints that induced immunity is leadings towards a herd immunity threshold in that the two earliest hit UTLAs (shown in black) had some of the weakest rises around the football spike, and now have some of the lowest case rates (cases/100k/day) in the whole of England.  If the stagnation of decay in cases was down to people becoming susceptible again, I wouldn't expect the plot to look like this.  Remember that these early to go UTLAs never spiked in case rates like the others because of intensive local health interventions on the ground - very important then as this new strain was more of an unknown and vaccination was less advanced.

So, not much to say this week, but all in all it still looks hopeful to me, and we'll see where we are with the ONS antibody survey when it comes out.

Once again thanks for all this. You and New Scientist are my go to now

I'm wondering if the graph of cases is as useful as it used to be or whether in England particularly people's attitude to testing has changed.  

https://coronavirus.data.gov.uk/details/testing

Shows the number of tests falling although not dramatically, but it is also a case of who is getting the tests i.e. how many are people who've been told to test x times a week and how many are people who think they might have caught it.   

If people aren't scared any more because they've been jagged and will be OK maybe they are less likely to get tested if the symptoms aren't too bad.

A changing relationship between symptomatic infection and testing comes up a lot as a potential explanation for big changes in cases.

Every time it ends up falling by the wayside as hospitalisations go on to confirm what we’ve seen in cases as real.  That’s unambiguously the case for the big “post Euros” drop in England and Scotland, and now with reference to plot 18 it’s also really clear for the much smaller, recent weather wobble.

I’ve never yet found my trust in the cases data to be misplaced - particularly the PCR only data for England (Scotland don’t release such granular data…?)

The most obvious reason for the number of tests dropping is that they’re symptomatically gated and infections are dropping….  Conditions conducive to covid spread are also going to spread the other respiratory viruses with overlapping symptoms…  The hospitalisations data and the gender ratio data both support this IMO.

I’m sure that there is a gradual disengagement from testing going on for reasons like those you posit but the key is that it’s not punctuate or sudden; it happens more slowly than changes in the rate constants due to policy, weather, an emerging variant and so on.

If you look in a frequency spectrum, I imagine engagement with testing is quite low frequency (< 10 cycles/year) along with gradual relaxation of rules, the weather effects are medium frequency (around 25 cycles/year), and stochastic noise is higher frequency.  Sudden changes of rules will ring through the spectrum.

I still trust cases as a useful barometer of short term change. The relative engagement with testing “drops out” of analysis like case hospitalisation or fatality rates if used to run forwards scenarios based on case numbers over the timescale of weeks - where testing engagement wouldn’t likely change much.

I'm pretty sure they do.   There's more desire for dropping restrictions in England than Scotland,  there are places in England which have resisted vaccination far more than anywhere in Scotland and there's been cumulatively far more infection in England.  I think England has decided the epidemic is over, Scotland hasn't.  Scotland is pretty happy with Sturgeon's caution, England seems to be OK with Johnson's risk taking.

I don't totally buy your frequency spectrum argument.   I suspect the 'freedom day' thing has changed attitudes on testing quite a lot and I don't see why it should have a similar 'time constant' to earlier events when the situation is different and less scary.  

I disagree and I have already laid out my reasons.  

I still find the cases data useful because both large and small fluctuations in cases still translate to hospitalisations.  

Further - I don’t think calculations of positivity support your stance on a sudden drop in the fraction of symptomatic people going for testing , rather testing fell because symptomatic infection fell as bourne out by hospitalisation data, and then the fall in testing stopped as the fall in infections stopped - again bourne out by hospitalisation data.

We’re moving to a world where testing should become irrelevant except for hospitalisations, and that has been discussed before on here.  Cases will be getting more detached from infection, that’s been acknowledged for some time here and seems both likely and a desirable indicator of exiting the pandemic phase and returning to normal.  But it doesn’t remove usefulness from the cases data in most ways, as the infections:testing rate seems to vary slowly and most of the interest and hence usefulness is in change, not absolute infection numbers.  

So, IMO, cases remain “useful”.  I don’t claim them to be a fixed ratio predictor of infections, and I don’t think engagement with testing was a sudden change following freedom day - I see no support for that in the data.  The main reason for the fall in testing was I think a fall in infection as confirmed by the hospital stats turning to decay with the right sort of lag.  I have no intention of getting dragged down your England vs Scotland sink hole, however I would note that I find the LFD data much more perturbed by sudden changes to testing regimes/engagement than the PCR data, and it’s a shame PHE manage to separate these for the public but PHS don’t.   

Change in the exponential rate constant from cases data can still be used to determine the direction of changes in the average daily temperature in England to something like 2-3 sigma.   The speaks volumes IMO…

Aarggh sorry I have lost the plot again - what is the difference between a "case" and an "infection". Huge sorries and I know this is obvious to everyone else.

An infection is when someone gets the virus.

A case is when they go for a test and get a positive result.

For the PCR testing, a positive case is almost certainly indicative of an infection.

PCR cases should be less than infections, and the ratio between the two probably changes over time - how likely are different variants to be symptomatic (hence sending someone for a test?), how do vaccines change how likely an individual is to by symptomatic and how likely they are to go for a test if symptomatic? 

Then there’s the bit where the number of tests carried out is (except at the crisis points) way, way higher than the number of positive cases detected; the symptoms of covid used to “gate” testing are very non specific so there’s effectively a random sampling effect to the testing as well (or even anti-correlated with some evidence that some colds have a prophylactic effect against covid infection…?). 

So, the relationship between actual infections and detected cases has always been a bit mysterious and fluid, but whenever thoughts have turned to changes in testing engagement as a possible explanation for big changes in cases, hospitalisation data has gone on to confirm a corresponding change in infection.

The ONS used to estimate new daily infections from their live infections survey - the two being not directly related but including the time an infection is live for.  They stopped; I do keep meaning to plot their live counts and the cases data together at least…

How do you work that out?  On your graph it looks like there are periods when they move together but sometimes they are wildly divergent or even go totally antiphase.

Shouldn't it be about whether you can predict changes in cases from the temperature not vice versa?

Also, how does the establishment of delta variant with a higher R show up on that plot?  Shouldn't that affect the relationship between temperature and rate constant?

Maths and statistics covered a few threads back.

Put both time series through a band pass filter to reject LF variation (variants, policy, seasonality) and HF variation (noise on a day-to-day basis), do a linear least squares fit between the rate constant and the temperature after the bandpass filter.  This shows a high R value for a correlation with a negative gradient (cases going up, temperature going down).

This correlation exists in a wide passband but does not exist in a null hypothesis using temperature data from a past year vs cases data for this year. The R value lands several standard deviations above the distribution of those arising from a bunch of  past year's temperature data.  So, P < 0.05.

There're a couple of periods where the effect is obliterated - these correspond  where we've had significant changes in policy or the new variant sweeping the old out etc.   

If there's a correlation, it's linking two variables.  So, given one, the other can be determined. "Predict" wasn't quite the word - the cases time series can determine what the temperature has done, without looking at the temperature.   With both that and the translation of those temperature wobbles to hospitalisations, cases still look "useful" to me.

As delta swept in some of the effects of that change fell in the same passband as the weather and that's one of the periods where the obvious temperature vs rate-coefficient anti-correlation weakens, the end of the football is another.  Even with these inseparable effects left in, the correlation still makes it through over the null hypothesis to statistical significance. Once the new variant stabilised, the effect returns - all sorts of things will probably change the scale factor of the correlation...

Quite right. England speaks with one united voice and its totally reasonable to characterise the nation as such. The Scots are definitely the best race. How would you rank other races?

The "with not of" thing is back on my mind. Tell me to shut up if you're fed up of it but I did the flaky sums again and, while it's nowhere near a big proportion, with cases homogenising through the ages, if they rise any more I'm thinking it'll be at the point where "within 28 days of a positive test" might need it to be taken into consideration. The withs don't have the same lag from cases as the ofs, and if (fairly big if) it's no longer a second order effect then that's soon going to start faintly messing with the relative delays in the rates.

Edit: forgot to say big thanks again for sticking with it and pulling these plots together.

It's definitely your version of "are we nearly there yet?"

I've updated my demographic projection (and spotted an error on the x-axis date labels, now fixed...).  As a reminder, this takes the by-age Covid cases data for England and applies an ONS all cause mortality model to estimate how many non-Covid deaths there would be within 28 days of a positive Covid test.  It doesn't include seasonality in the all-cause mortality which means it probably under-estimates deaths in winter and over-estimates them in summer.  

The ratio has improved dramatically (like a 10 fold) since vaccination started, but it's now hovering at around 11 (probably higher if I had seasonality in my mortality model).

I would make this a core weekly plot, but it's a bit to poor methodologically (lacking seasonality) to really be fair I think, still indicatively - no, we're not there yet.

Agree; if we can reach "with" then the rates would have no lag; as it is perhaps OOM 10% of the lag is seemingly reduced by the "withs".  

We've both talked before about likely ongoing disengagement of some from testing - I think the people more likely to continue engaging with testing are the ones more likely to have serious health consequences, and testing remains a feature for suspected cases admitted to hospital, so I''m not sure we'll ever get to the point you're thinking of with cases, because at that point the problem is boxed up and symptomatic testing starts to become an anachronism - we either expand it to all circulating viruses or taper it down and focus on random surveys... It's nice to dream...

It currently represents about 10% of my "all subjects weekly plotting output" so it falls within the procrastinatory budget... 

Edit: I really wish I'd had this plot late last summer when we had more rallying cries along "run over by a bus with Covid" lines..

And there was I thinking it was about a 10-12% effect.... Reassuring that I can guesstimate the same figures, although I think I basically did the same maths but only for a single date. Not sure what it needs to be before it starts to muddy the analysis but it's one to remember when headscratchy things start to happen.

"Houston man shot 6 times last week still waiting for surgery at hospital overwhelmed by COVID-19"

This is what let it rip or anti-vax means in a  developed country.

https://www.fox26houston.com/news/houston-man-shot-6-times-last-week-still-...

A picture says a thousand words.  I got some good feedback from minimike (A walking version of Boas?)  when I showed my workings on this, as the bandpass filter I was using was prone to ringing.  I think we agreed this makes it very sensitive to the signal, but also to the effects of noise.  Or maybe we didn't agree.

Anyhow, I'm now using a bandpass filter made in a way that's easier to describe to life scientists and that doesn't ring so much - details on past threads, but the pass band is around (0.3 ± 0.2) cycles/week - so variations over the timescale of ~2-3 weeks fall through it.  

Plot A top half breaks the rate constant signal in to that to the low side of the bandpass (black line), that in the bandpass (grey shading) and the high side of the bandpass (purple shading).

Plot A bottom half shows the signal in the pass band for the rate constant (left y-axis, black line) and the signal in the pass band for the temperature (right y-axis, red line).   The rate constant signal's value is equivalent to the height of the grey shading from the black line in the top plot.

The values in the pass band can be considered as showing the short term deviation from the low frequency baseline signal, excluding the worst of the noise.  So a positive value means "it's a bit warmer today than it has been or will be for a while" or "the rate constant is higher than it has been or will be for a while".  

The shading on the bottom plot is blue when the pass band values for the two measures are of opposite sign (warmer than local average, rate constant lower than local average, and vice verse) and red shows when they're moving in the same direction.

This pretty much just pulls out to stark clarity what I see by eye when looking at the plot in my OP, but it does so in a testable way.

How to show this is a real effect?  Plot B makes the same plot, but using temperature data from other years.  This is the best I can find for a "null hypothesis" - the seasonal trends and the general ranges of temperatures and temporal power spectra will be drawn from the same population over recent years, but all the details of timing will be different.  Plot B shows the same coloured plot swapping in temperature data from 6 recent, past years.  It's clear from a quick glance that there's a lot more red - about half and half, as one would expect from random data.

Using a statistical test to compare the findings from Plot A against the null hypothesis embodied in plot B I get about 3.5 sigma certainty.  So, more than a 99.7% probability that the effect is real.  This would not be enough for a detection in particle physics but a pretty good significance threshold for the life sciences.

I’m pretty sure people in England and Scotland don’t think all that differently .

good work- you must have enough for paper/book now? just needs a storyline...

Thanks.

I should probably write up the temperature stuff and put it out there, scientifically speaking.  I’ve walked away from the insipid negativity of peer review (very different experiences in different fields) but once more won’t kill me.

”How to make plots and alienate people”?

“A plot too far”?

”Chumbawamba and the rate constant”?

I’m flattered, but no. I cant hold a candle to the great MB.
 

Excellent..

Can you make a plot that shows phase angle rather than crude sign? The binarisation of the result could be misleading..

indeed! Science!!

Hmm, yes but waves are just collections of coherent particles, so I’m after 5 sigma Thankyou very much ;-p

What's all this now?? I thought the gold standard for declaring a finding on UKC was dividing one number by another?

I can and will, but the laptop isn’t coming out again now until tomorrow so I’ll do it then.  The 3+ sigma significance is present in the R value of a linear regression between the curves if that helps allay your concerns.  The shading on this plot was more to show the trend than to be scientific.

I don’t think it’s particularly valid to expect the magnitude of the correlation coefficient to remain constant against changing seasons, vaccination and variants however…. 

Isnt it great how there’s whole accepted ways of testing a hypothesis in a meaningful way that transcends pointing at numbers and shouting?  Science is great.

Well, if the pandemic goes on for another 50 years or so, you just might get it….   

Friday Night Covid Plotting #2639:

Evolution of the scientific method:

killing people->shouting->pointing and shouting->pointing at numbers and shouting->pointing at numbers->scientifically robust hypothesis generation and testing->hacking p-values->pointing at p-values and shouting->pointing and shouting->shouting->killing people

where are we today?

I ran out of energy for the garden chores, so put the plots together.

We are somewhere in the part I put in bold, and will probably be towards the end of the above once Trolly-McTrollface catches up with us.  

Plot C - scatter plot of data from the lower half of Plot A, with quadrants coloured according to the colour of the shading on lower Plot A.  The legend gives the LSQ fit parameters and R value.

Plot D - the same plot performed for the years forming the null hypothesis shown in plot B.

Plot E - a box plot and a mean ± standard deviation for the R values from the null hypothesis years, and a data marker for the 2021 R value - null hypothesis violation is ~3.8 standard deviations.

I'd welcome your view (and other esteemed readers who work in the life sciences) on how appropriate this is as a null hypothesis method; well outside my usual area of analysis.

Would this convince people it's a real effect?  I think the violation on fraction of datapoint in anti-phase is going to be higher, but is a bit more of a qualitative observation - but without an understanding of the causal mechanism linking the variables there's no reason to expect the gradient of the linear fit to be time invariant, and probably lots of reasons to expect it to change with all the other stuff going on...

Someone forgot their leap years!

Guilty as charged 

I buy it.. as an observation of correlation. What I struggle with is the causation..

Well, that's somebody else's problem!

LSRH had pondered if it was to do with sample degradation in vehicles in high temperatures, but the effect persists in hold and cold seasons of the year, and it shows in the hospital admissions data.

Given that most transmission happens indoors, it strikes me as likely behavioural.  

There's a similarity between what the bandpass filter is doing and what the time derivative of each data set would do (which also shows a striking correlation) - it's looking at change in the signals (d/dt for derivative - going up or down, variation from local mean for the bandpass - above or below typical - not so far removed from each other).  When things are nicer, there's less Covid spread, when things are less nice, there is more spread.   This feels strongly behavioural to me - people dress for the season - filtering out the seasonality form some of their behaviour, then adjust their behaviour to the daily weather - if it's nice, more likely to do more things outdoors or open more office and bus windows etc.  If it's cold, more likely to withdraw indoors more.   I know  a few people who wear trousers for 50% of the year and shorts for the other 50%, then make their preference for beer garden or inside based on if its actually shorts weather....

Or: "I have found the causal mechanism but this journal's page limit is not sufficient to explain it."

Maybe covid virus particles are a potent cloud-seeding agent..............

I think there was more mileage in my snot production theory than the hot van theory, because at least that could plausibly affect transmission, but neither really holds up across seasons quite like the "I'm cold, I'm going inside" theory does.

I've been waiting for someone to suggest that 2020 should be excluded from the null hypothesis, because it's a Covid year.  A causal link from Covid cases to the weather isn't totally batshit insane, as there were dramatic changes to road transport and civil aviation patterns, and aviation in particular measurably affects planetary albedo.  I'd love to do a parallel-universes test of this and peer review - exclude 2020 on these grounds and one wold probably be labelled a lunatic, include it and not mention the very low probability of conflation and one might get called up on it...

'snot a bad theory, but with the latest wobble being visible in hospital data (I need to do a 7-day rate constant method on the rest and see how much previous wobbles sow through there, just visible easily this time because it changed the sign of admissions) I think that rules it out, because there'll be a test on admissions that's carried out in a temperature controlled environment?

OK.

But when you look at your graph, the period of time where the curves have the most obvious relationship i.e. both rising together is actually the opposite of what your theory predicts.  It is a coincidence that temperature is rising because of the season while Delta pushes up R.

I don't have time, or the skills and inclination to get into the maths in detail.  This looks superficially like the kind of thing I was trying a few years back modelling cryptocurrency markets and trying to predict whether a coin would rise or fall.   Same kind of stuff - either do it by hand with filters and stats or use a neural net or learning algorithm and let it figure it out itself.

It all worked pretty well on the training data, just like your model fits your few months of data because it was constructed it from that data and tweaked until it fit.  But it was useless for predicting what will happen in the future and fairly useless if applied to a completely different time period it hadn't been trained on.  Unlike Covid there are a good few years of historical data from cryptocurrency markets to test the models on so it was possible to train/develop the model on one year and test it on a different year.

Covid looks to me to have similar properties to turbulent markets - the thing mutates, vaccines come in, mass events like football happen, politicians make stupid decisions, school terms happen, stuff happens in other countries and it drowns out small correlations.  I'm not saying there is no correlation between temperature and Covid spread, that's a completely reasonable hypothesis but is it a large enough factor to be useful in predicting the future or does it just get drowned out by random events.

The effect is so obvious that tweaking is not necessary.  Using the time derivative of both measures (no free parameters, no tweaking) produces comparable results.  The range of pass bands over which this holds is remarkably wide.

But, critically, the correlation coefficient achieved doing this is way bigger than the best that can be achieved using null hypothesis data drawn from any past year.  This is the critical core of the analysis that makes it statistically quantified science and shows to scientists that it is not over training a model (or p-hacking an analysis) on limited data.

Perspective adjustment: It’s not a predictive theory, it’s an observation of a relationship in the data endowed with a high statistical certainty of being real.  Someone else can come up with the theory to explain why it is there.

In terms of being the opposite of what I’m talking about - it is not.  I’m saying that the data supports an anti correlation between the two variables on the timescale of a couple of weeks, the “obvious relationship” you refer to falls over much slower frequencies.   It’s the deviations from that gradually changing baseline that anti correlate to high statistical certainty.

Yes, the anti-correlation I have shown to exist with high statistical certainty exists over timescales of a couple of weeks.  It seems to be coincidence that the emergence of delta and the shift from winter to summer mean temperature occurred at a similar time leading to a seperate correlation in the data over a much slower timescale. 

Both these effects (delta and winter>summer) occur slowly compared to the pseudo-cyclical variations in temperature and rate constant, which can be shown to have a relationship on sorter timescales that is highly unlikely to be due to statistical noise.

The anti-correlation in the (wide) passband holds despite these slower effects.  I can see that by eye clearly in my first post in this thread.  Look at the short term deviations from an imagined pair of trendlines - print it out, grab your flexible edge and draw them.  That’s basically what my analysis is doing.  One goes over, the other goes under.  

Yes, statistical methods applied to a bandpass filter centred on timescales typical of the short term variations in our weather show a high probability that the anti correlation of temperature and cases exponential rate constant on the timescale of weeks is real. 

There is no neural net or learning algorithm here.  I like my data analysis old school and thoroughly understandable.

This is a statistically quantified observation on the data.  I first made it about 3 months ago (using d/stuff rather than bandpass but the maths is not dissimilar) and it has basically held since then - as shown by a null hypothesis test.

This is not a “trained model”.  This is an observation of a correlation in the data with a statistical test.

I feel that it is a little bit insulting for me to say this, but cryptocurrency is not like Covid.  At all.

The underlying mechanics are not so similar.  Actually, they’re almost polar opposites at times.

In particular the nature of feedback in finance is often proactive and positive (faith or panic goes on to reinforce more faith or panic) where-as the nature of feedback with covid is reactive and negative (cases have risen and hospitals are filling up, shit we need to fix that).  This makes one system fundamentally more prone to chaotic effects (in a mathematical sense) and the other system more damped and chaos proof.  Beyond that, the value of currency (crypto or otherwise) is basically unconstrained, where as the mechanics of covid are thoroughly constrained by population size and human interaction/contact rates.

Again, cryptocurrency is not like covid.

Solid scientific methods put a high certainty (> 99.7%) on there being a link between temporal mesoscale variations in temperature and the exponential rate constant for covid.

If you want to challenge that, you need to poke holes in the analysis, which IMO your post has not done.

The anti correlation of changes in temperature and rate constant has proved remarkably predictive since I first noted it 3 months ago.  Other things also happen as you note; most of them happen on shorter or longer timescales than the temperature changes so it seems they’re pretty separable.

Is it predictive going forwards?  It is until it isn’t.  If we get a source of change with a similar characteristic frequency and a similar of larger magnitude, that breaks the link for a while.  Perhaps someone will identify the causal link, and effect policy change to break it giving us the lower possible rate constant going forwards.  Still, it hasn’t let me down yet…

I don’t really care if it’s predictive or not. I care that it can be evidenced to a high statistical certainty in the last 8 months of data as being a real correlation in the data.  I care about this because it shows how a measurable (outdoors temperature) affects another measurable (rate of covid spread, largely happening indoors) when the outdoors/indoors split should mean there is no connection.  This implied the existence of causal links that, if understood, offer cutting insight in to what to do differently indoors to lower the rate constant.  This likely has value to the winter flu seasons as well.

I feel you may have got a bit hung up on the word “predictive”.  What I meant was that the time series cases data alone can be used to deduce fluctuations in the time series temperature data leading up to those cases.  I have shown that to a statistical certainty of > 99.7%.  That this relationship holds cuts back to your first question: (i paraphrase) “are cases still useful?”.  Yes, and this temperature relationship is like the double red underscoring on the real reasons why they’re useful.  (They still predict changes in hospitalisations and deaths).

Edit: fixing typos.

I mean.. it IS. And it used a neural network. But the optimiser wasn’t called Adam and the feature space was strongly regularised.

 What did I miss?

Any model is predictive if someone applies it to the future.. although it may not be any good in that scenario.

When it comes down to it, everything ever computed can be boiled down to a bunch of polynomials, which is all that ANNs boil down to.

But the analysis is in no way a case of throwing a software ANN/MLP/RNN at the data and getting an understanding-devoid result out.   In a pragmatic sense I mean the analysis is rooted in mathematics and statistics we understand, not in a bunch of automatically refined random numbers and sigmoids giving an answer that we’re left scratching our heads about…

If you’re implying my input in to the process can be boiled down to a (sufficiently large and fast) computational equivalent to a recurrent ANN I think that’s a ballsy interdisciplinary claim to make right now, and Gödel would like to have a chat with you…  Possibly also Maxwel and Boltzmann.  So I feel your theoretical grounds for your statement are as yet unproven.  It’s all the rage to state that consciousness will be an emergent property of a sufficiently big ANN but so far nobody has actually constructed one in the form of a testable hypothesis…  One of the great mysteries to be answered in our lifetimes?  I’m not sure the process of getting the answer will make the world better, mind you.

This is true, and as I said it’s predictive until it isn’t.  Meaning much the same thing.

But what I mean is that I care far more about the insight it hints at (in 40 foot burning letters) rather than predicting the rate constant next week.  

I’m teasing, as you know ;-p. You say random sigmoids, I say basis set! You say brain, I say neural network. You say intelligence, I say regularisation. 

science is built on (a near infinitesimal subset of) ballsy interdisciplinary claims. 

I’d give my eye teeth to be able to chat with Godel.

Looks like we’re in to your final stages (pointing and shouting->shouting->killing people)

Do you remember the big fuss when SpiNNaker launched? One might idly wonder “where’s the triumphant follow through?”  I’m more or less convinced we should be trying to design a truly analogue substrate with significant thermal noise as an alternative approach.  The next decade is make of break for the current direction of research…

Near the top of my list.  James Clerk Maxwell and Oliver Heaviside are above Gödel because I’d have at least a small chance of keeping up with those two…

We’re not shouting. We’re hypothesis generating. Completely different stage of evolution. ;-p

You are missing my point.

The difficulty with cryptocurrency and - I suspect - Covid is that there are extraneous random effects arriving that drown out the correlations that you've observed.

In crypto, over the range of time for which there are historic data for e.g. BTC / USD or BTC/ETH or the one I liked to toy with BTC/DOGE a lot of other stuff happened.  Tech booms, financial crashes, wars, changes of US government, futures market starting up,  invention of new blockchain algorithms, frauds being discovered and if you were doing it this year a pandemic.   You can look at the market say BTC/DOGE and you can find a reasonable theory to explain/predict the dynamics of based on other financial/market data you have.  Just like looking at temperature vs Covid cases.   If you work it over 8 months of historical data you can get pretty good results - either doing the sort of stuff you were doing or with machine learning / neural networks.

Then some extraneous macro-economic sh*t nothing to do with the effect you observed hits the fan and your model fails and you either stop using it or lose a ton of money.  It isn't because the effect that you'd figured out has gone away it is because some much larger effect has come along and in the current market circumstances it isn't important.  

I think Covid is like that.   Three different strains so far, large changes in government policies about isolation and travel, testing, vaccination etc.  Every few months the world changes the rules: this is the similarity with crypto markets.

If you had the temperature vs cases model and were trying to make predictions of new cases for most of the last 8 months you'd be saying: mmm I don't know because .  There's always something new, unquantified and potentially larger.

No, no I am not.

You are missing the point.

The observed effects are not “drowned out”, they are shown by statistical methods to be real to a high degree of certainty.  They aren’t the only effects, that is obvious.  But they are there and they are demonstrably real.

Hint: Coming back with a load of utterly unrelated stuff about Bitcoin does not invalidate an analysis of covid data.  

Yet, despite all that the temperature anti correlation has persisted for the last 8 months (the only time period with PCR only data, important for consistency).

Once again you spectacularly miss the point. 

A large faction of the variance in the rate constant for cases over timescales of a couple of weeks is explainable by fluctuations in temperature.  I have shown a null hypothesis to put a probability on this.

Unrelated waffly bollocks about Bitcoin has FA to do with this.

Have you tried reading what I’ve written?  Let me repeat myself.

I don’t really care if it’s predictive or not. I care that it can be evidenced to a high statistical certainty in the last 8 months of data as being a real correlation in the data.  I care about this because it shows how a measurable (outdoors temperature) affects another measurable (rate of covid spread, largely happening indoors) when the outdoors/indoors split should mean there is no connection.  This implied the existence of causal links that, if understood, offer cutting insight in to what to do differently indoors to lower the rate constant.  This likely has value to the winter flu seasons as well.

Further, in terms of “always something new” - yes, yes there is.  But often it falls at different characteristic frequencies than the weather correlation it seems.

Well, no, because I’m not dumb enough to over-fit a model to some currency data without performing a statistically quantified null hypothesis test and without performing a sensitivity analysis on my model, and then to put money on it.

A couple of identifiable effects have upset the anti correlation in temperature.  Even with those data points left in the ensemble still has high statistical probability of being real.  For sure other events will upset it for brief periods again.  But I’m not interested in prediction, rather understanding how factors affect the spread of covid.  Because the goal here is not to make money gambling on predictions but to understand what makes covid spread, so we can find smarter, less invasive restrictions to make things better.

I’m going to close out this discussion from my end.  I’ve talked through the details of a scientific method with a scientific, statistical test and you’re not picking holes in any of it, just drawing totally inappropriate analogies to cryptocurrency which had fundamentally different mechanics.  

There are many factors affecting the spread of covid.  Some are obvious, some are not.  I am working to evidence one that was less obvious.  I think I have evidenced it’s presence to a high scientific standard.  Yes, lots of other things modulate the rate as well.  These factors are not mutually exclusive.  

Edit: you seem stuck in a bit of a rut so I’ll repeat it again.  I’m not looking to predict the future but to evidence the impact of one unexpected variable on the behaviour of cases - mesoscale fluctuations in temperature.  Lots of stuff we don’t anticipate is going to happen, but at least now I understand one more of the factors affecting what happens.  If we stopped trying to understand anything because other things might happen, no progress would be made.

I’ve moved beyond shouting and am waiting for the capacitor bank to charge.  Going to be waiting for a while yet…

Why would the indoor/outdoor split mean there is no connection?  If it is warm outside people spend more time outside and they leave windows open more often when they are inside when the heating is off.  People walk or use a bike instead of a car or public transport.  Doesn't seem at all surprising that if it gets a bit warmer there would be less spread of Covid.

It would be surprising if it was linear with temperature over an extended range if it is about how humans react to weather rather than weather itself.

You've only got data from Jan to half of August of one year on your graph.   Do all the maths you like it is still not that much data.  You've only seen one winter and one summer.

Your statistical test doesn't tell you much about the limits of your theory.  All you can say is that for the way the weather was in 2021 in the UK your hypothesis looks OK.  If next year was much hotter or much colder or much rainier maybe people would behave differently and the correlation would be lost.

I wonder if you tried the analysis of temp vs case data on several US states with significantly different climate and different types of buildings it would work for all of them.

My stats are remarkably weak, but doesn't the null hypothesis test show that there is bugger all possibility that this effect has happened by chance? So isn't the amount of data is perfectly sufficient?

If I had found that sort of edge on bitcoin, I would t be wasting my time posting on forums, I'd be trading myself to financial freedom.

So in my thicky brain (I can't follow the maths), it seems like there's an identifiable effect linked to the weather, and this allows policy and procedures to be adapted, and then their effectiveness could be measured. That would seem a useful thing to me.

What I'd still love to do, and I don't know how because there's no equivalent data set, is this analysis against humidity as well as temperature. Just seems like something that needs ruling out but not sure how to go about that.

I agree, and I think WT should be encouraged to publish his findings in the academic literature (seriously)

seems your thicky brain isn’t so shabby Bridgstarr.. 😉

I think most important thing you can do now, is use your model along with the government's model for cases in the areas around key climbing locations and accurately predict the weather for the next 6 months.

In particular, if the case rate remains as it is, will I need a waterproof at Stanage on Saturday and what time will the car park be full.

Especially as in brexit Britain we are told we are  in sunny uplands..

Roy Lilley has been focused on government pressure on NHS comms during covid.

"See what she does...

Usually, it’s the result of perseverance, sacrifice, learning, study, the love of what you do and a sprinkle of ‘smarts’. How to measure it? The achievement of objectives, goals or maybe, simply peace of mind? Success is not what you have or don’t have. Success is what you think about what you have, or don’t have. And, often, it’s what we think others think about us, that decides how content we are. It’s all about perception… and perception is everything. If it walks like a duck, it’s a duck… perceptions. Perception is reality. What people perceive to be reality, is their reality and there’s no argument.

Governments live and political careers die by measurements of perception. The measurements are called ratings. HMG, BoJo have to be perceived as successful, on-top-of-things. The reality for the people, the voters, is public opinion and that must be cultivated and curated, at all costs. Comms people call it, ‘the optics’… how it looks.

When success doesn’t come as a result of perseverance, sacrifice, learning, study, the love of what you do and a sprinkle of ‘smarts’, the temptation is to engineer it. Massage the numbers, fiddle the facts, spin and ultimately, bully people.

During the pandemic it was important, for public reassurance, that HMG portrayed itself as on top of things, in charge and competent. It wasn’t. Mistakes were made, decisions delayed and science was learning. It used its Covid-powers to take over messaging to the public and that included an iron control over what the NHS said. That cascaded into the restrictions imposed on communications professionals in our Trusts. The over interpretation of that role, by NHSE, has lead to the avalanche of grief and angst that has overloaded my inbox. The stories of pressure, abuse, and outright bullying are still arriving.

As one NHS boss puts it; 

‘… as a Chief Executive there’s always some degree of watchful eye on what we may say in the media, but nothing like this. It sounds almost like state censorship at a time when local communities really needed to know what was happening in their local NHS.’

The NHSE press office, faced with a dilemma, could have said to comms leads; ‘We’re in a difficult situation here, No10 is adamant they want to control messaging, to avoid misunderstandings and public alarm… how can you help us deal with it…’

Or… 

… well, we all know what they did instead and are still doing… in the words of an NHS comms professional;

‘…I was part of those regional Comms teams spending fruitless hours marking the homework of excellent Comms people. I spent time in tears not believing in what I was doing and exhausted…’

The optics are bad.

NHSE could still say;

‘What’s come to light is not what we intended, we were all under pressure and we got a lot wrong. We are sorry. Thank you for what you did, can we start again…’

I can't imagine it... can you?

The common image of the ‘workplace-bullied’ is a weak, oddball, loner. The academic evidence contradicts that. The bullied are often successful people, well-liked and professional. Bullies pick on capable people. They’ll try and pick off the influencers because they see them as a threat. Bullies pick on people who are better than them, know more, have more experience and expertise because bullies are generally out of their depth. They bully because they can’t, or don’t have the skills, to achieve their goals by leadership, management skills, persuasion, communication, experience or talent. So, they can end up white collar thugs.

As one emailer put it; 

‘Sometimes, they made it feel like we were working in a totalitarian state.’

How common is bullying in the NHS? NHS Employers tell us;

19% of staff experienced bullying or harassment from colleagues

13% experienced bullying from a manager

98% experience incivility in the workplace

Toxic behaviour costs the NHS £2bn year.

The Guardian report bullying and sexual harassment, 'endemic' in NHS hospitals.

Workplace relationships are set by the tone at the top. Attitudes resonate downwards and what is imitated, copied, licensed and happens at the front-line, starts on the desk of the boss.

Although new in her role, all of this happened on Amanda Pritchard’s watch, as COO.  

It will be interesting to see what she does…"

Yes.  Did you read where I said “This feels strongly behavioural to me - people dress for the season - filtering out the seasonality form some of their behaviour, then adjust their behaviour to the daily weather - if it's nice, more likely to do more things outdoors or open more office and bus windows etc.  If it's cold, more likely to withdraw indoors more. ”

I get a general impression you’re not reading much of this and replying to individual bits out of context.  It might explain an all to familiar feeling I’m getting about the direction of this conversation…

Funny, you seem to have been having a very hard time accepting a scientific method showing that this is the case.  I was surprised that the signal comes through so clearly in the data.  Finding the *right* way to measure the rate constant from very noisy data was the key step.

I have not claimed that it was linear with temperature.  I have shown that the sign of the effect is anti-correlated with the sign of the local *deviation* in temperature.  I have shown that a linear fit to the local deviations has statistical significance, but this neither implies nor proves that the effect is linear, and not do I claim that.  It is however irrefutable proof that the link is there in the data.  Someone else can come up with a theory on the mechanism and what sort of form the relationship should have, and can test that.  Looking at the dispersion to the linear fit however, trying to get scientific tests of quality of fit on higher order models is probably not a great use of time…

What is your point?  You expend a lot of rage and two threads trying to prove which vaccine is better by dividing one number by another.  When those numbers had a rider saying “do not do that thing Tom is doing, it’s not valid”.  “Doing all the maths I want” is showing that the apparent strong anti correlation in these two datasets is > 99.7% likely to be real and not due to chance.  I don’t need to do any more maths to show that, that is a gold standard for discovery in the life sciences.

Do you have an evidenced basis to claim this “one winter and one summer” are somehow extraordinary or not representative?  Also, you forgot about spring.  Seems to me we’ve covered a typical set of temperatures now.

I have said this before.  It is not a theory, it is an observation proved to high certainty on the data.  Someone else can come up with a theory to explain it and gather measurements to see if it fits or not.

I’m sorry, I missed the part where I had been claiming anything else.  Previous discussions have covered other obvious factors that are likely to be involved but are hard to incorporate.

The hypothesis proved to a gold standard level is that over the last 8 months, when it’s getting warmer, the rate constant is dropping, and vice versa.  It doesn’t look “OK”.  It looks 3.8 sigma.

What matters is finding the causal mechanism, and exploiting it as much as possible specifically so that this effect is *not* visible going forwards.

The papers shared a couple of weeks ago on seasonality from international analysis supports you here I think.

The data you seek is out there, in the BMS logs of managed facilities and in the servers behind things like Google Nest.  I think you’d need to be a heavyweight epidemiology group with connections to be able to open doors to see if you could find a way to access anonymised data.  There may be buildings models linking inside and outside conditions to people’s houses etc…?  Or crowd sourcing a Raspberry Pi data collection across homes and offices?  Quite the project but absolutely worth exploring.  Modelling is probably the way forwards; beyond hobby level stuff.

In reply to minimike:

It’s on the list.  Moved up it a lot getting the plots with the new bandpass together for this thread.  

In reply to Philip:

Obviously I have predicted that already, but if I release the information too soon, too many plans will be made on it and the predication could be invalidated.  So I’ll post it about 8 pm on Friday night…

Yep, there's just no equivalent as easy to grab as the hadcet temperature.

Loads of papers out there about this general thing that support what you're doing and the conclusions it leads to. These are some I found this morning:

https://www.nature.com/articles/s41598-018-37481-y

https://www.annualreviews.org/doi/10.1146/annurev-virology-012420-022445?ur...

https://academic.oup.com/cid/article/49/6/861/333844

I once asked our met office for humidity stats in order to try to work out how tightly I should pack wooden floorboards. They sent me a graph that looked like a year of continuous earthquakes with no obvious seasonality. The explanation was basically that when the wind had any sort of west in it and brought in air from over large expanses of sea the humidity rocketed, when it was easterly and mostly came over land it was drier.  As it can be high humidity without actually raining (or threatening to) I don't think it makes a huge difference to behaviour. A better thing to try to correlate to would be weather forecasts. 

In reply to Wintertree:

Can you see any 'end of month' artifacts in the data/graphs?  It's quite noticeable here that town is significantly busier on the end of month weekends when people have been paid compared to early or mid-month weekends.

Wasn't suggesting humidity affects behaviour

I like your thinking, I get my Met Office contact to run some fake forecasts for bad and then good weather, and we see if there're corresponding changes in the rate constant!  I see no ethical or practical difficulties ahead...

It's true though, people make plans based on the forecasts as much as the mach one eyeball'd nowcast; then often doggedly stick with them as conditions change...

Is the "Payday Bender" still a thing?  I suspect it's a tall order to look for, as the noise is quite bad, and a single-day event get smeared out over a few days of this rate constant measurement.  I mean, I can try...  Update:  I had a look at the "residuals" in the high frequency bad - the part of the signal not in the low frequency band of weather pass band.  Plot below.   There's no obvious day-of-week effects to the noise; the biggest peaks are the anomalies from bank holidays and the football I think; some of the wobble is probably the most rapidly changing part of the temperature data to get encoded in cases, but rejecting it also rejects the unrelated noise such as the bank holiday effects and some of the football.  

But it may make a difference to the lifetime of viral particles, or to the viral load in exhale etc.  Lots of biophysics going on there.

That being said - high humidity compounds the heat, leaving me with an urge to do no more than sit on the terrace with a cool drink...  The idea of socialising seems far to much like hard work on such days...

In reply to Longsufferingropeholder:

Thanks for these - some to add to the list.

I think the short term link is a really nice addition to the seasonality studies. Covid is I think going to lead to some long standing questions and debates over "seasonal" illness being settled or at least radically redefined.  Just imagine if they do need to happen cyclically due to the virus/host dynamics, but with the right nudge we could anchor their bursts to summer, not winter... 

In reply to thread:

Update 7-day rate constant plots below.  England is hovering very close to a fixed daily rate of cases, Scotland is in growth with quite noisy measures - not exceptional levels of noise compared to past data, and not unexpected with their lower population.  The temperature anti-correlation in the medium term changes to the rate constant continues to hold up quite well in England.  The weather looks to be in a holding pattern for the next week, with some talk of a late season heatwave to follow...?

The stats say it hasn't happened by chance.   They don't say it will persist for a different range of temperature values or that it wouldn't be drowned out by some other, larger factor which was not present in the time period  e.g. suppose next year the temperature was a fair bit hotter or there was a lot more rain.

It isn't that hard to train an algorithm on historical data and find some statistically valid link between price and some other variable over a period of 8 months.

The problem is the conditions for which the effect you found is larger than other factors do not persist.  A few years ago I created some code that was pretty good at predicting which way DOGE would move relative to BTC.   I was fairly happy and trying it on really small trades for a month it was all good.  Then the whole market crashed and everything was falling like a stone. 

The effect my code modelled was suddenly small compared with forces it didn't know anything about.  It was obviously time to chuck it in and go do something else. 

My point is that 8 months of Covid cases vs temperature is like 8 months of historical data on DOGE/BTC.  There's valid relationships you can find in that data.  But are they going to still be useful determiners of cases  if e.g. it was significantly warmer or significantly colder, or significantly wetter, or there was a new strain of Covid next year.

Yes, it is useful but it's also useful to know what range of conditions it is valid for so you know when to use it.  We obviously can't get historic data for a longer time period for Covid but we could look in different geographic areas with different climates.  That's why I suggested doing the same analysis for a few US states.

And yet...

• We have the football effect (demonstrable by the gender statistics) falling mostly in a different (higher) frequency band.
• We have the rise of the new variant falling mostly in a different (lower) frequency band.
• We have confirmation that this effect persists when the baseline temperature is cold, and when it is hot.  
• We don't expect wildly different temperature ranges in England than those seen over the last 8 month, and rumour has it there's been a lot of rain at times.  

I'm not claiming this effect will persist, so I'm not sure why you're stuck in a broken record mode on making totally inappropriate and irrelevant analogies about your obviously flawed Bitcoin get-rich-quick scheme and understanding the spread of a virus whose mechanic has absolutely bugger all at all in common with the mechanics underlining currency let alone cryptocurrency.

A bloody big asteroid could hit the planet next week and end the virus, along with all human life breaking the link.  That Marbug outbreak could go airborne and force us all to properly self isolate, breaking the link.  None of the reasons you've given seem so likely given that this sort of thing has already happened, and the frequency spectrum analysis of the data does a nice job of spectating quite a few of them.  I'm sure theme will come where other effects once again fall in the weather pass band and conflate the data.  So what?  Part of the skill in analysing this data is getting to the bottom of conflating factors, understanding them and - where possible - separating then.  The more of those factors are understood, the easier it is to identify the unknown ones.

Also...  I feel I have mentioned this multiple times and you continue hell bent on ignoring it, so please excuse the bold face but I am not interested in prediction, I am interested in understanding what has come to pass and what can be learnt from that about the wider mechanics of the spread of the virus.

I both disagree quite strongly with your bitcoin analogy, and your wider take.

Point.  Head.  Woooooooossh.

Great idea, you could go do that and report back to class.

Bit rich to dismiss >3 sigma as happenstance after last week's discussion of analysis methodology.

You can read about similar studies looking at similar, and different, viruses. I linked a few above.
This is definitely a thing.

Yes, the disconnect is just staggering.  My brain has nearly imploded under the weight of the irony.

Yeah, I wrote a rather more verbose piece about the astonishing incongruity but then pared it back to a more simple wtf, then eventually settled on that.
Anyway, found tons more papers about connection between respiratory viruses and temp/humidity. Have skimmed a few and read even fewer. All have tons of refs. Browser tabs demonstrating how exponential growth works. Don't go down this rabbit hole.

I have not dismissed anything.  I have not said there is no effect in the data, quite the opposite.   I've said there's only 8 months of data and given an example where a valid relationship discovered in one year's data was irrelevant in another year.   

I got nothing. This is like watching Alanis Morrissette trying to put out a burning fire extinguisher.

Sounds like we’ve had the same ear worm this evening.   A traffic jam…

In reply to tom_in_edinburgh:

It’s not a valid example.  Because - and honestly it feels more embarrassing every time I say this - because covid is not like Bitcoin.  Most of the “but this may happen” reasons you think it might suddenly fail have already happened in the last 8 months.  By happenstance they fall in different frequency bands.  

You still seem obsessed with saying this relationship may not hold in the future.  I don’t recall ever claiming it will a as I said, “it will till it won’t”.  I care about what it can teach us about the last 8 months.

I suspect it will continue to hold because other effects are likely to fall in different frequency bands, but I’m not claiming to have proven this.

You’ve banged on about over fitting a model before. Ignoring the silliness of calling this analysis a model, do you know how many free parameters it has?  3.  Do you know how many data points there are?  241.    Getting a correlation of 3.8 sigma significance with 80x as many data points as free parameters is not over fitting or trawling through every possibility to find one that fits.  You can tell it’s a really good analysis because that ratio is really large :-o

Edit: fixed a typo; ‘t’ and ‘r’ are right next to each other on the keyboard…

I find your theory interesting, if I didn't find it interesting I wouldn't have wasted my time thinking about it.  When I find things interesting I try and pull them apart and find weaknesses and limitations, my own ideas as much as anyone else's. 

You didn't claim it could be used to make predictions but I am interested in whether it could be used to make predictions.  My interest isn't the same as yours and I ask about stuff that interests me.  

You don't need to sell me on the statistics and correlation.  I'm happy to accept that for the 8 months of data that you have you've got very strong correlation.  

Which makes me wonder why you've spent so much time arguing against its predictive ability...

Ah, I see my mistake.  I didn't recognise this as a question: "You've only got data from Jan to half of August of one year on your graph.   Do all the maths you like it is still not that much data.  You've only seen one winter and one summer."

The way you have repeatedly said why you don't think this is long term predictive comes down to two areas:

1. Other things will change
2. You couldn't predict a crypto-currency's future value

I dismiss (2) almost instantly as a joke shop argument although one part is worth addressing. You've suggested you could train a model to match the data well in your bitcoin case, but that it didn't predict.  How many degrees of freedom did your model have, and how many data points?  This analysis finds a strong correlation with 3 DOFs and ~240 data points.   (2 DOFs really as there's some degeneracy in them and it boils down to the two -3 dB points of the bandpass filter).   You have repeatedly suggest any model could be fit, but the quality of fit obtained with 3 DOFs against a couple of hundred data points shows that to not be a valid concern here in terms of how these things are evaluated scientifically.  That is the only lesson I think that could be carried forwards from your cryptocurrency.  As a scientist, when I decide if there is merit in an analysis or model, it's based on questions like "How many DOFs does it have vs data points?".

In terms of (1), I have re-cast the data from Plot A upthread in a different way.  I thought plot A would get the point across, but I'll spell it out to help answer your question on if its predictive.

 This shows the rate constant signals in the low (L), medium (M) and high (H) pass bands. 

• L - this is where the effects of policy, variants, emerging high levels of population immunity and perhaps seasonality show through.  These are slowly changing effect.  It's clear from looking at the top rate constant plot, that this is the main predictor of the rate constant, and therefore cases.
• M - this is the passband where the analysis finds a statistically strong and numerically large correlation with the weather.  In terms of predicting things, it's not of great use because we can't predict the weather that far in advance, and because it's an AC signal - the net modulation of the rate constant over more than a few weeks is zero, so the cumulative effect in total number of cases is not so large although it shifts their timing around.  (Edit:  The exponential mechanic translating this in to cases is non-linear, meaning that a balanced, transient event in this AC signal does still have an effect on the total number of cases after its passing, but for small deviations compared to the "L" signal, the effect is small) 
• H - higher frequency signals.  Some noise effects and some "dragons" - fast acting events encoded in the spread of the virus like the Euros football finale.  

If I wanted to predict in any meaningful way what's going to happen, the "L" band is where I would focus my attention.

Understanding the "M" band for me has two benefits:

1. For my weekly interpretation, it gives some insight in to recent changes in the rate constant.  If it was changing not in concordance with the observed models to date, that would say "something has changed" and that might be something good (football ending) or something bad (a new variant rising to prominence so rapidly that it lands in the M band not the L band).
2. A big flashing neon sign saying "There's really interesting epidemiology to be done". 

Not everyone is paid at the same time though. I had assumed that the rise in positive tests reported Wednesday to Friday each week, was a result of people mixing more the previous weekend.

That's an invitation to come up with a counter argument.

As a result you showed me the graphs L / M / H graphs that I'd not seen before and I have a better understanding of what you are doing and why you are confident.

I actually could, obviously not perfectly but enough, and the market conditions it could deal with persisted for longer than the amount of time you have data for Covid.

https://www.coingecko.com/en/coins/dogecoin/btc#panel

You need to click 'Max' to get the full time line.  It's not about data points and degrees of freedom it is about falling off the edge of the known world.

It predicted OK in 'normal times' but didn't during a market crash.  I don't think that is unusual.

Obviously the crypto market is a very different thing but I'm only trying to make one simple point.  When you've only seen 8 months you've not seen everything that can happen and it is hard to know the limits of applicability of what you have done.

I think your temperature vs cases hypothesis could have limits of applicability.  For example, if it gets sufficiently hot people will start staying in more when the temperature rises rather than going outside more when the temperature rises.    Similarly, I would guess there is a threshold below which people aren't going to spend much more time outside when the temperature rises a little because it is still far too cold to be outside.

What we have here is a failure to communicate.  Reviewing the messages, I think I've been pretty clear...

As I had never claimed that it would predict the future you might perhaps understand why I am a little bit confused by all of this...  You've presented these without explaining the context that you are challenging me on something I have not said, and all I have done is explained why I consider the observation over those 8 months valid, and why I think your specific reasons it my not apply going forwards are either nonsense (cryptocurrency) or evidenced as having fallen in other frequency bands to date (weather stuff).

It never occurred to me that you were apparently assuming I thought this was predictive, as I've specifically explained multiple times and shown in graphs it applies to a relationship to temperature fluctuations over the time scale of a few weeks and that's not the sort of thing that is well forecastable beyond a couple of weeks out, so does not lend itself to prediction...

It is however a phenomenal tool for interpretation of the now, and for spotting if the now is behaving egregiously differently to expected trends, and for hinting at fascinating epidemiology to be done. 

I thought from previous discussions on other topics you'd have got that pretty well from the term "band pass", and Plot A, and the plain English description.  Anyhow, now there's a graph that spells it out, always a good thing.

Cryptocurrency is not like Covid.  I really don't know how to make this any clearer.  Do you know how many epidemiologists use the fluctuations of financial markets as a way of interpreting pandemic data?  I'll be it's a number somewhere between zero and zero - because the underlying mechanics of the two are totally different.

You seem confused.  In an earlier post, you said:

It all worked pretty well on the training data, just like your model fits your few months of data because it was constructed it from that data and tweaked until it fit. 

That is precisely and exactly talking about degrees of freedom (tweaked) and number of data points.  I put the relevant bits in bold. 

I challenge you to fit a model to 240 days of a cryptocurrency's value that has only 3 degrees of freedom.  You can't, and do you know why?  Because... currency fluctuations and the mechanics at play with Covid and with the weather are completely different.

I'm rather confused about your myopic repetition of this followed by a recent retcon that you were questioning me on it when it's not something I've claimed, and it's something I've agreed with endlessly - that this link is not guaranteed to remain indefinitely in the future (prediction is a red herring as to me this isn't predictive beyond the immediate future) although your reasons why don't seem compelling to me, there is always the unknown.  In particular I pointed out that most of your potential reasons fall in different frequency bands, and the effects seems to superpose in the rate constant - which is not surprising when you think about the mechanics at work.

Speaking of repeating myself.  It's not a hypothesis it's a statistically quantified observation on past data.

A hypothesis could be a theory to explain why the correlation is there, or it could be to postulate that the effect will remain visible going forwards.  Such  hypothesis will obviously have limits of applicability.  There is no "limit of applicability" to a quantified observation of some data.  That is an analysis set in stone, whose bounds are defined by statistical tests saying how likely it is to be due to fluke rather than some causal link.

Yes, not an unreasonable suggestion unlike the extraneous waffle about your failure to predict cryptocurrency values.  However, the last 8 months have seen a pretty wide range of temperatures for England, so unless we get some truly apocalyptic weather this is pretty improbable.  Still, it is a good jumping off point to explain one of the differences between a cryptocurrency fluctuation and weather....  The weather is bounded, and fluctuates around long term seasonal means, with some gradual drift in the means over the years.  Temperature returns to near the mean after every big heatwave or cold spell.  Cyrptocurrency value is not bounded, and is not held by physical forces to only limited duration deviations from its seasonal mean.  If the equivalent of your "market crash" came to the weather on a short enough timescale to fall into the "M" band in this analysis, and it didn't recover towards the seasonal mean, we'd all be dead of a new and horrific ice age or global firestorm. Think before you ask these questions, Mitch.

Tell me why I can't fit the data between Feb 2016 and Feb 2017 with three degrees of freedom.  I can draw a straight line through that on the month to month time frame.  If this was Feb 2017 and Feb 2016 - Feb 2017 was all the data available it would pass any stats test your data passes.  

https://www.coingecko.com/en/coins/dogecoin/btc#panel   (and click MAX)

Tell that to somebody who is short on DOGE in April 2017. 

The stats are just as valid as before but it is irrelevant because the world just kicked the table over.  That's the problem with 8 months of data.  You've very likely not seen everything that can happen, there are far more likely to be 'unknown unknowns' than if you had a decade or two of data.

The analogy to a 'market crash' is not a weather based model failing because we got an ice age, it is a weather based model not being useful because everyone is so effectively vaccinated it can't spread anyway or because the virus just mutated to something much more infectious.  

It feels like you’re still missing the key point that I’m not dealing with long term trends here.

Cryptocurrency is not like Covid.  I put a graph of your data below so others can see what I mean.  Looks to me like it tanks in value then some random event causes a near instant spike then common sense kicks in again.  Utterly different to covid or weather, no physical bounds on rate or range.

I am not fitting a model to a time series, I am evidencing a correlation between two time series.

Which, as I’ve said until I’m blue in the face, fall in different frequency bands.  Hitting herd immunity thresholds did not break this temperature correlation.  The rise to prominence of each of Kent and India has taken longer than a typical period of temperature variation.  Herd immunity emerging is a slower process than that typical period.

A “market crash” is sudden; the kinds of changes you posit as breaking this going forwards (a) have already happens and not broken it and (b) are unlikely to fall in the same frequency band as the effect and (c) are market crashes which are not strictly bounded in the way the spread of the virus and the changes in temperature are.

Why on earth would I do that?  You asked if I could fit a 3 DOF model and I showed I could fit a linear model.  If I've got a linear model that fits the largest trend in the data then I'm done, I don't filter it out and look for something more complex model that explains smaller, faster fluctuations.

There's a return to mean effect in markets and there are practical constraints on how much money could flow in but I agree a low cap crypto coin is very volatile. 

The final spike is insane even for crypto.  I think Elon Musk said something about selling cars for DOGE.  The other bumps are more typical. 

Right.  And actually, the crash didn't break the effect I was looking at either.  What it did was make it uninteresting from a practical perspective because there were much larger effects at play.   You are welcome to say from you perspective you don't care whether the temperature effect is a significant factor in the overall number of cases, you are only interested whether it is a real effect. 

I have a practical/pragmatic perspective and for me it is extremely important whether something is one of the largest effects which influence the number of cases.  I don't much care if the largest effects are in a lower frequency range - a low frequency effect can almost completely drown out a higher frequency one.

I don't think you've understood what this is about.

I thought it was pretty obvious I was talking about fitting to the medium frequency pass band but I’m learning to spell everything out…

What I have endlessly and repeatedly said is that it’s an *important* effect because it tells us there are mechanisms affecting the spread of the virus that we don’t understand.  Perhaps if we can understand and control them, then the effect can become larger.

I have not been talking about an effect in terms of the “number of cases” but of the rate constant.  Small changes to a rate constant have big effects over time.

I don’t think you understand something pretty central to this.

Regardless of the "entertainment" being provided by you two, the H/M/L graphs would definitely be needed if this was to be presented to people like politicians rather than those with some scientific abilities.

I'm sure you have plenty of brick walls around the house, although junior might wonder what daddy is doing 😁

Yes, a tangible improvement in how I present the method/results - something useful for my time.

With the state our walls are in, I have to limit myself to the paving slabs out back.  Although today I have something far less pleasant to occupy my time - I was 6 months over-due a trip to the dentists when the pandemic landed, so today I am going to be in The Chair for some time, catching up on 30 months of dentistry...

Edit: The HML graph with the linear fit to the temperature drawn in red on the M graph.  Clearly not perfect, but also very far from uncorrelated.

I'm pretty sure I do.

What I'm saying is it is time to move on from stats showing the effect is in the 8 months of data you have.   That box is ticked.  I don't care if you put more 'walls round it' because you've already sold me. 

I'm now on the next step: 'OK so there's a correlation but is it useful to take action based on it.'  That kicks off a completely different set of questions. 

This is a cool graphic but right now when M spikes up it looks like the top one is spiking down so its not illustrating R = L + M + H.  Maybe you should flip the M graph so when your eye adds the bottom curves its obvious the result is the top one.

Hmm

Welcome to square #1

Agghhhh.... please let's change the subject ...... How about maskgate as the difference between government and opposition benches is just so stark (with a few honourable exceptions).

https://www.theguardian.com/politics/2021/aug/18/ministers-accused-of-putti...

Good call. You have a guardian link for every situation

Since you raise it though, the ones that did did so "while not speaking". Implying they took them off when they did their tw*ttish, loud, pontificatory shouting across the room. Brilliant. Pissing directly into the wind there then. And Starmer didn't get a personal mention that I can see there like he did in every other publication. Nevertheless the guardian has told us to be angry about it, so upset we must get.

Interesting...

Guardian: "All the Labour frontbenchers, including the party leader, Keir Starmer, and his deputy, Angela Rayner, wore masks."

Independent: "However, Sir Keir Starmer, Labour’s leader, did not." 

Ok so.... Eh?

Sky: "Ahead of making a contribution, some MPs and party leaders wearing a mask would remove the covering several minutes before speaking."

Great. So overall difference made: 2/3 of cock all then.

The point is they (those with masks) followed the advice given by the speaker. The others didn't. It's pretty bad for half the house including all the front bench to brazenly ignore the advice given. (Especially in the age of 'personal responsibility'!) I agree with you it would be better if the kept them on while speaking, but it's not remotely true to say that if they take them off for speaking then there is no point wearing them at all as you implied. They are in the chamber for several hours and on average only speak for a couple of minute, if at all.

I also agree with Offwidth that it's time to change the subject!

Also, your anti Guardian stuff is a bit tiresome but you can read the BBC instead on this one:

BBC News - Covid: Most Conservative MPs ditch masks as Commons returns

https://www.bbc.co.uk/news/uk-politics-58259604

More interesting and less political findings in this study:

https://www.ndm.ox.ac.uk/covid-19/covid-19-infection-survey/results/new-stu...

Some of the findings are on subjects we have recently discussed and some are counter to the received wisdom in the past, quoted below:

Two doses of either vaccine still provided at least the same level of protection as having had COVID-19 before through natural infection; people who had been vaccinated after already being infected with COVID-19 had even more protection than vaccinated individuals who had not had COVID-19 before.

Delta infections after two vaccine doses had similar peak levels of virus to those in unvaccinated people; with the Alpha variant, peak virus levels in those infected post-vaccination were much lower.

Two doses of Pfizer-BioNTech have greater initial effectiveness against new COVID-19 infections*, but this declines faster compared with two doses of Oxford-AstraZeneca. Results suggest that after four to five months effectiveness of these two vaccines would be similar – however, long-term effects need to be studied.

(Edit - * the study text also says "There was no evidence that the effectiveness of two ChAdOx1 vaccinations >14 days previously in preventing new PCR-positives differed from the protection afforded by previous natural infection without vaccination (heterogeneity p=0.33), whereas two BNT162b2 vaccinations afforded greater protection (p=0.04).")

"The time between doses does not affect effectiveness in preventing new infections, but younger people have even more protection from vaccination than older people.

Within the COVID-19 Infection Survey, recruiting representative households across the UK general population, SARS-CoV-2-associated symptoms varied by viral variant, vaccination status and demographics. However, differences are modest and do not currently warrant large-scale changes to targeted testing approaches.

I read a few different outlets' takes on it. The spin in each is depressingly predictable.

I'm still of a mind that if they actually gave a crap about spreading virus they wouldn't spend 5 hours crowded into a stuffy room when they've proved they can all tell lies at each other just as effectively over the internet. An unspecified, badly fitted bit of cloth over their faces is doing nowt in that situation. And the amount of virus that comes out when talking is going to be, basically, most of it, so I'm sticking with 'very little point', other than virtue signalling. They'd have stayed outside unless contributing if they actually wanted to help.

Yep, saw this came out. Need to read it properly but I can already see how the imminent 'better jag' pissing contest is going to play out here.

I'm afraid I think that came out of the daily mail spin box.

MPs can't do their job without going to parliament. The government effectively set the rules about attendance (virtual or not) through Rees-Mogg and have subsequently ignored the advice given by the speaker about masks. The opposition have a choice to either go and wear masks (or not), or not go and not represent their constituents in an important debate. To claim that going to do their job while following the applicable advice is 'virtual signalling' is BS.

The stuff you say about the mask effectiveness is unsubstantiated and patently wrong. Two things can be said about mask wearing with absolute certainty: (1) they will reduce viral spread, (2) we don't know how much. Any other assertions about it not making difference (or the opposite) are unsubstantiated. In the context of them having minimal impact on the wearer when seated, it is clearly proportionate to expect someone to wear one, which is why such advice still applies elsewhere, and why it applies in parliament.

I knew you would bite

Starmer took his mask off, under the agreed H&S code, while he was speaking.  The point of the article was a voluntary code on masks was put in place as part of a covid  H&S response in an indoor workplace. A lot of such H&S plans allow people when speaking not to wear masks where communication is important from funerals to Parliament.  Some MPs took the code seriously but too many others didn't, sadly very much along party lines, so given the extra risk to all workers in Parliament, maybe the code needs a review.

Masks make a difference if the wearer is infected, risks are higher when population infection levels are high (which they are) and MPs have more and more widespread contacts than average members of the public in work. It seems important to me from a H&S perspective they wear masks (unless they have exemptions) when not speaking in the chamber. That's before we think about more general points, like the government setting a good public health example for their own messaging about masks to the nation.

But hey, why would our government rendering their own workplace H&S precautions ineffective and giving a terrible public health example in direct conflict with their own public messaging be important when a good rant at a guardian link is available?

If we must talk about Tom's latest confusion on this technical point: we are way beyond the position that it became obvious that both 'the pupil and the teachers' were not doing their jobs properly (Tom isn't trying hard enough, or is trolling, as its unlikely he is incapable of understanding and others seem more interesting in putting him down than educating respectfully). I've had occasional revelations myself where despite being pretty good at academic stuff I'd inadvertently got stuck in an incorrect position of understanding, but a good teacher (or recognising something seemed 'off', so trying to work out what that might be) usually got me out of it. It helps me have sympathy when people 'don't get' stuff that is normally grasped quickly by others. Banging on and on about something that someone clearly doesn't understand in front of peers is arguably 'dumber' than the misunderstanding.

Some data on this in the study:

Vaccine effectiveness was also generally higher at younger ages (Table S3). For example, VE 14 days after the second BNT162b2 dose was 90% (85-93%) for those aged 18-34 years versus 77% (65-85%) for those aged 35-64 years (heterogeneity p=0.0001); and was 73% (65-80%) versus 54% (40-65%), respectively, for ChAdOx1 (heterogeneity p=0.002).

The advice to government that public indoor mask use has been assessed as the second most effective measure of stopping covid spread after vaccines has been linked on these threads several times. It's long been clear that won't change the more sceptical view on mask effectiveness of some regulars.

I take your general point (as I understand it) from past threads that masks aren't a magic bullet and the guidance needs to be clear so they don't give people a false sense of security, but this is really far more fundamental - the team blue MPs are showing contempt for the control measures introduced in their workplace to protect all employees.  Directly undermining messaging / public support.  It's the exact opposite of what I expect from government.  

In reply to Si dH:

Some reading for me later today.  

Sounds like there's no support for the fringe idea that getting one's first dose of antibodies from live infection is "better" then....  

I'm more or less resigned to getting infected with Delta once Jr goes back to school; I eagerly anticipate the trial results on using Valneva as a 3rd "booster" jab given its major difference to the other vaccines, and assuming that's good, it would be nice to get a dose of that before delta...  We heard on another thread from one of the long time workers on the vaccine program that they're gearing up for boosters, but without knowledge of which product...

Bruxist shared a paper a few weeks out (out of McGill?) with a detailed study.  For something like a lecture theatre - and by extension the commons - you'd want masks and decent mechanical air handling.  An indoor room occupied for hours on end is towards the worse end of the spectrum for spread.  Re: ventilation - presumably another way in which Westminster is probably antiquated and not fit for purpose....?

I do take LSRH's point (as I've understood it before) however that if someone feels safer than they are from mask wearing, this could have counter-productive behavioural effects.  But I suspect the bigger problems are in the venues where mask wearing is non-sensical such as pubs and restaurants. 

In reply to Offwidth:

I tried my best to find an angle from which to get the key points across.  I shouldn't have assumed that they understood what a bandpass filter is or would make it clear if they didn't... Got there in the end with the H/M/L plot...  Much of the discussion Tom is interested in was I think covered a few threads back - where someone was posting but was at the time more interested in painting JCVI as an English plot...   If there's a clear failure to communicate going on I always welcome a 3rd party stepping in to try and explain the issues from a clear external perspective...

It's just not true. We've been over this. That 'second most effective' crap only holds if you caveat it with 'apart from staying outside, distancing, not going to crowded places, and all the other much more effective things.

Irt Si & wt:

I've painted the wrong picture of my attitude to all this. 

I'm in favour of face coverings. Wintertree has summarised my feelings better than I did. They help. If they're a properly fitted one made to some standard, they can help significantly. But if you're in a stuffy debating chamber for 5 hours near someone who has covid, you're getting covid. If you act as if a shabby facepants makes you invincible, you're definitely getting covid.

Yes, the govt should have followed their own advice. They made an arse of themselves. 

Yes, the biggest sigh I will sigh today came out when I saw Offwidth had posted a link to a guardian article about the government making an arse of themselves. We'd almost got through a whole week...

Your H/M/L plot is very helpful in explaining what you are trying to do.  If you write a paper on this you definitely want that in it.

Of course I know what a band pass filter is.  It's the sort of thing you pick up when you have 35 years experience in the electronics industry.

I think it's more (some of) the blue team place a higher value on not being coerced (as they see it) into behaviour than not infecting others with a virus they see as largely controlled.  I think it's more disagreement than contempt, which is different.

I always admire your efforts but sometimes you need to learn when to stop. Help isn't useful if it's obvious the basic principles are missing (or stubbornly ignored). However, I'm more annoyed with others who insult more readily.

I've always been interested in the general area from doing mathematics that had fantastic real-world application, in the Laplace and Fourier I studied early as a physics undergrad,  and a control final year option that had the theory but was more focussed on interesting real world case studies, like US motorway traffic control, and where the limits were and why some approaches could risk things going badly wrong. I taught aspects of introductory control and filters to engineers for decades, when I took over various modules that colleagues were not queuing up to teach.  I usually turned them from one of the least popular to one of the more popular, doing my best to follow on from the inspired teaching I'd had. My main aims, alongside the nuts and bolts, was to inspire with clever applications, alongside highlighting failures, to instil a precautionary principle. Placement students, research colleagues, spin-off companies and personal contacts exposed me to more industrial practice. However, as things went more digital I became less of my physics background and more the engineer.  The added complexity of sampling and digital filtering in fourier space sometimes even gave me 'brain ache'. Relatively simple Matlab system modelling outputs away from generic designs could still surprise me at times (until I thought the issue through very carefully), let alone real world implementations. Generic solutions that worked in a stable way were usually the best choice

As we have discussed before...... adding things that need effort and experience to understand, to the modern ease of building highly complex MIMO systems in safety critical environments, sometimes with inappropriate unstable interfaces, increasingly linked to the web, and with some dishonest companies who would blame the human link to disguise a bad design or security..... and I worry a lot. On the plus side of digital modernity it was a joy to use design tools in MATLAB and not have to inflict ordinary engineering students with some of the tortuous mathematical constructs of the analogue days.

Which is why I was surprised it took the H/M/L plot to get the penny to properly drop.  I had assumed it would be obvious for you that the content of the M band is both high and low pass filtered, with the high pass filter removing any DC component, so that - even if we had a prefect forecast of the weather for the next year, the next effect would be close to zero, as the mean value added to the rate constant from this band tends to zero over longer timescales, pretty much by definition.

Hence my utter confusion that your repeat instances it wouldn't be predictive - with which I readily agreed - and when you kept pushing on this.  It's a medium frequency band - it's an irrelevance beyond the parameters of that band.

Bullshit. They have a public health message on mask use indoors for good scientific reasons and should be providing a positive example for that. Infection levels in the community are high and an MPs 'connectivity' is much higher than average. A significant proportion of the double jabbed can still catch delta and spread it and LFTs usually pick up infections a few days too late, if at all.

The non-mask wearing majority of backbench tory MPs are simply putting opinions on personal freedom before the science, as many of them always wanted to. The real scandal is ministers are doing the same.

Yes we have been over it. It's indoor advice for people who can't avoid being indoors and  that includes mask wearing being more effective than distancing indoors. 

Your 'shabby facepants' rants are in my view stupid anti-scientific undermining of an important public health message. No one claims a magic bullet but you seem to be ignoring the combined clear benefits (even if well below 100% of a well fitted hospital mask) as spread is probabiliistic and linked to viral load ingested.  I linked some surprising success stories before including examples where infected individuals wearing fairly standard masks didn't spread covid despite massive indoor contact. There is even a documented case where mask wearing highly anti-correlated with catching  covid  in an environment where it was circulating (the US warship).

That's exactly what I said.  You are so busy being the internet hardman shouting BS, you don't read what you are replying to.

Actually I read it and it's  bs. We are back to a previous disagrement on the subject. You can't have a valid opinion that directly contradicts the science. Their disagreement in that context is a semantic way of disguising contempt for the scientific advice for the H&S of all those working in Parliament. That's as individual MPs, the damage to the public health message is a deliberate choice of our government.

There's an important distinction between what I'm calling a mask and what I'm calling facepants. We spent a while going over that one paper, remember.

And then there's taking it off when they shout over a crowd of people's heads. That's as socially responsible as wearing glasses but taking them off to drive.

There's surprising stories for everything. There's not great literature on face coverings but what there is, correct me if I'm wrong here, dues not claim they're more effective than staying the f*** outside. Not all of them spoke. They did not all need to be in there. That was arguably one person's (jrm's) call, but not really.

If I'm honest I'm still struggling to get over your post a few months ago when you had a sanctimonious rant about people not wearing face coverings on a train when you'd got on a train to go to the cinema. I don't think we'll ever meet in the middle about the lengths we each think one should go to to avoid indoor spaces. One person's can't avoid it is another person's really don't need to. There's always a choice, but the question is at what cost, and I'd set the balance in a very different place than you seem to.

Ref: original point about many MPs being knobs for not following guidance/courtesy, yes, I agree, that was another massive PR own goal.

Actually you didn't refer to backbench specifically. The main problem as Offwidth says is the frontbench. The advice to wear masks in crowded spaces etc is their own guidance, so how can they possibly disagree with it? The only two possible scenarios are (1) they agree that their guidance should apply in the Commons but decided not to follow it anyway or (2) they don't think it should because one rule for them, one for the rest of us...etc. Both paint a very poor picture.

Valneva vaccine - UK initial order (Sept 2020) specifies 60M in late 2021.
Maybe soon, maybe on 31/12/21, maybe later. 

They are recruiting heavily - "Vaccine company Valneva creates over 200 new positions at West Lothian facility" The Scotsman 16/8/21

Phase 3 trials are not yet complete.

Anybody seen anything more concrete?

https://valneva.com/press-release/valneva-announces-major-covid-19-vaccine-...

https://www.scotsman.com/health/covid-scotland-vaccine-company-valneva-crea...

https://valneva.com/press-release/valneva-completes-phase-3-trial-recruitme...

I just laughed at it.......as it will be a very public loss of face if alot of them have to isolate.......

It could be incredibly embarrassing for the Hof C and that is the best way of looking at it.

Give it a few days.

Not sure what you mean by 'been over'. The paper is clear in its conclusions, facecoverings  are a major benefit marginally above ventilation and above distancing but below vaccination.

We agree being outside is by far the best way of reducing risk (again I've consistently argued outdoor restrictions were more about behaviour risk rather than covid risk for people distancing; and sometimes questionable given the mental and physical health benefits of going outside and meeting others safely outside). Yet a large proportion of our population has no choice but to use public transport and/or have to work indoors.  Most need to shop for basics and some sometimes need to help a friend or relative not in a bubble..

Something as significant as the US warship covid data doesn't happen by chance. By far the most likely explanation is mask use protecting the wearer and/or indicating peculiarly significant different risk behaviour for wearers.

As someone who totally believes in the science around covid I've been way more sympathetic to people not following government rules and advice  than most here. Poor messaging, dishonest information on the internet and peer pressure are bad combination before we introduce booze.  My comments on the train were not sanctimoniously aimed, they were just reporting what happened. Yes my risk threshold is clearly much higher than yours. I was double jabbed as was my partner (who tests regularly for work) and neither of us was due to meet anyone indoors in the subsequent week. I'm shocked at your shock as a fellow climber who deals with deliberately chosen risk

As for the MPs I don't care about PR anything like as much as I do about PH

It's more your scientism is so deep you don't seem to be able to imagine any other way of seeing the world, which is probably why you spend a lot of time raging impotently at those who do see things in ways other than scientific.

That's a more reasonable point about hypocrisy but the discussion was about the "blue team" generally, the frontbench hasn't come in to it until now.

Why do you feel the need to make such childish personal attacks? Of course I can see how others see the world. I'm an agnostic with many different religious and atheistic close friends and family. I'm an ex academic scientist/engineer who has worked closely with other academics with very contrasting views. I've regularly defended good people with religious views on UKC from blanket attacks from our resident militant atheists. I've even urged kinder treatment from some regulars to the odd covid denying poster who I knew was a genuine climber (most recently trad dad)  and more generally urged understanding of broadly good people who break covid rules because of misinformation and poor messaging. What is likely to get a strong reaction from me is people who claim to be on the side of science being wrong or misleading.

People are entitled to believe in stuff that is completely unscientific but that shouldn't ever take such precedence that a government behaves in a way that completely undermines their own public health message. On an individual level, most of those non-mask wearing MPs have claimed to be following the science.

No, it didn't say that. It said a good mask with an added clippy thing over the nose was about as good as good ventilation, and a facepants is nowhere near close. Still a benefit, don't get me wrong, they're a good idea and an easy win and wear one, but they're just not that great compared to expectations.

This is where I like to argue the phraseology, not necessarily the point. It's too easy to say there's no choice. It can be a shitty choice, an obvious choice, someone else's choice, Hobson's choice, but in the overwhelming majority of cases it's a choice. For sure some of the time working indoors had to happen, but we could have done far more to make different choices that would have reduced the amount of time spent indoors with others by making some harder choices. I'm not saying we should have in all cases, but I am saying we should have in more cases than we did. Saying "can't avoid it" is a cop out. Saying "what would it take to avoid it?" and then deciding not to because it comes at far too great a cost is a better way to put it. Certainly makes employers think a bit harder. And it's where I think our MPs should have cone to their limited senses.

I'm not going to get into this. There will be an equivalent anecdote somewhere on the internet where some bunch of people in masks caught more covid. We all agree they have some effect and we should keep on with them but it really needs more thorough studies. Still a benefit, don't get me wrong, they're a good idea and an easy win and wear one, but they're just not that great compared to expectations.

Have we gone all the way round again yet?

Your wrong on the paper it was a realistic look at real world mask use. It was open that the quality of mask use and ventilation varies.

Being in an indoor space with people from other households really wasn't a choice for many. It maybe should have been for a significant proportion but it wasn't.

The US warship isn't anecdote it was an accidental research experiment.

https://www.sandiegouniontribune.com/news/military/story/2020-06-09/cdc-stu...

As I understand it:

1. Phase 1/2 results - a very positive press release but no scientific publication (?)
2. Phase 3 - in progress, we await the results....
3. Manufacturing plant - build out in progress, will require inspection and licensing before product from it is used, separate to the approval of the vaccine itself.
4. A new trial is underway to evaluate efficacy as a 3rd "booster" jab, as well as a separate trial to determine serum level responses from older adults.

A lot of the usual tight lipped secrecy over commercial issues.  As perhaps the recent FT article (linked on #38) on Novavax shows, it's not guaranteed to be plain sailing in terms of manufacturing approval but I've got a good feeling about Valneva.  I don't think it's just wishful thinking...

Given the large investment by the UK government in this plant and the nature of the most recent trials, it's tempting to speculate about where this is all going, but the timescales probably don't work out for getting the most vulnerable a third dose in time for the winter.

I'm sorry to see bad-tempered posts proliferating on this thread, with its pedigree of reasoned, polite and informative debate.

If I might interject on a less politically-charged matter..... has anyone noticed the cases by specimen date for Scotland for the last few days?

https://www.travellingtabby.com/scotland-coronavirus-tracker/  (you have to scroll down a little)

If the last three days of figures, and particularly Aug 16th, are not anomalous, then growth has become quite rapid here. The timing is not consistent with the return to school (which was either yesterday or the day before for most pupils), nor does it seem to reflect recent weather (along the lines of wintertree's excellent analysis above). Instead it is consistent with an impact from the final relaxation of restrictions, including the opening of night clubs, on Aug 9th. I have heard reports locally of a significant letting down of hair by some youths in the aftermath, and indeed the growth in cases does seem to be fastest amongst the nightclubbing ages, and the pattern fairly consistent across the Scottish council areas.

Pure speculation of course, and one that can only fit with the history of English cases if one accepts that there is lower infection-based immunity in Scotland (which is not unlikely). If that is the case, then England can watch on with interest. If, however, Scotland is still a couple of weeks ahead of England in a common unfolding of a disease dynamic....

I await the release of today's data with some trepidation.

I'm not. It doesn't support your claim at all, anywhere.

Yes it was.

It was a choice that someone, somewhere made. Not necessarily by the individual, but it was always a choice. You can't just generalise and say lots of people had "no choice". You can say it's a prohibitively costly and impractical choice, but as soon as you put it in terms of the alternative it immediately forces a bit more justification from employers etc. And in so many cases things could have been done differently.

Today's data have arrived. Ouch.

I think one thing that we're seeing writ large in a lot of this is what conservatives really mean when they talk about valuing "personal responsibility" as opposed to rules and regulations. They tend to want us to believe that it's about trusting people to look out for their neighbors based on common decency without needing some petty rules to force them to do, but when they're actually given the opportunity to display some "personal responsibility" themselves it turns out to mean that people like them can do whatever suits them and screw everyone else.

Funnily enough I'd just been having this conversation offline as well...

Working theory - and one we've seen before in England - is that the sudden rise in cases is due to the "switch-on" of LFD testing, which would start picking up symptom free infections that already existed before school started.

The increase in exponential growth rate again with today's data isn't so consistent with that.

Without full demographic data on cases and in particular without the breakdown by PCR and LFD that PHE gives but not PHS, it's hard to say.  If it's a "switch on" effect it would not endure in rate constants, but just as it fades out, school associated transmission might feed in.

That's just one working theory thought, it could be something else less palatable...

Several of my daughter's school friends caught it in the last week after going to a party in someone's house and on to a nightclub.   They were seeing their friends before going off to Uni.   Glad she was sensible and decided not to go.   That age group have mostly had one jag and are only a couple of weeks away from the second.    

Lower infection based immunity seems like the most reasonable explanation for Scotland getting strong growth where England remains flat despite Scotland keeping some restrictions.

It's a pity the nightclub restrictions didn't remain in force until the teens and early twenties had got both their jags.

There's definitely a lot of that, basically selfishness and hypocrisy.  There is also genuine concern about loss of freedom in the longer term and some kicking back against that  Not entirely unfounded. E.g

https://www.bbc.co.uk/news/uk-scotland-scotland-politics-58244323

The notes page for the Scottish dashboard (link below) suggests the published results exclude LFTs. LFTs increasing in number as schools go back could still cause a rise in the case numbers published as they are confirmed by PCR, if they were high enough in number. Seems possibly far fetched though.

The dashboard also includes a positivity graph. I only have my phone with me and can't see it well, but it seems to be going up?

https://public.tableau.com/app/profile/phs.covid.19/viz/COVID-19DailyDashbo...

Yes, nightclubs + lower infection based immunity could be part of the picture.  The demographics on Travelling Tabby suggest 15-20 and 20-25 years of age are the biggest growth areas.  

Could be a mix of school associated LFDs (is this happening north of the border with the return to schools?) and nightclubs...   If it is some mix of these, hopefully very little translation in to hospital admissions...

Edit: At Wicamoi's suggestion from a week or two back, I added a second red line to the Scotland rate constant graph for nightclub and similar restrictions being dropped; this is coming in at the right kind of delay to be compatible with those restrictions going as a source of additional spread.

Thanks for your thoughts.

The reminder of the potential confusion that changes in case:infection ratios, consequent on testing regime, can introduce reminds me of a suggestion I had for you. I think I have previously expressed my surprise at the strength of the correlation between a crude measure of temperature and case rate constants - it makes me think something else may be going on than just animal behaviour (even when amplified by human broadcast weather forecasting, social planning capacity, and availability of unusually airtight and shareable shelters). If I were reviewing the paper that I hope you'll write on your temperature analysis, I would be thinking about possible artefacts - not in the analysis, but in the data itself. Might it be worth conducting a co-analysis of ‘tests’ rather than ‘cases’? Temperature-related differences in the infection-to-case relationship could potentially amplify (or mask) the effect you've identified. For example, good weather might reduce hypochondriacal feelings, so that fewer asymptomatic, or near-asymptomatic infections are tested/detected when the sun shines.

I guess we can all think of multiple ways in which both amplification and masking might be favoured, and I don’t expect this would reveal anything significant, but any rate you could satisfy reviewers that the phenomenon is not explained as an artefact of variation in testing rates, or at least put some bounds on the potential size of such an artefact. As I say, just a thought, and one that previous experience leads me to believe you will have already considered about 3 light years in advance!

I think the apparent uptick in rate of increase arrived too early for it to be school-related testing. Many teachers went back on the 16th (major rise), but not pupils.

Positivity rate has indeed been rising in Scotland

Good suggestion and easy enough to implement.  

Given the typical low positivity rates in good times, I still wonder how much the symptomatic testing is actually symptom-free infection being detected when symptoms of something else gate someone in to PCR testing, making it effectively random (or a mix of random and even with anti-correlated effects given that some colds are shown to block out infection by Covid...).  So, weather can affect the non-Covid symptoms of asymptomatic people being gated in to testing.  Loads of possible effects, hard to wrap my head around the full range of what to control for.

The other part of this is looking for corresponding signals in the hospitalisation signal; it stood out this time because the rate constant oscillated around zero, it makes me think the relationship is also likely there for other temperature wobbles in an all-positive and all-negative regions.  That's a very strong confirmation if it's there that this is related to infections not just cases.

All good ideas - I think going so far as possible to make the case that it's not just a real correlation, but really indicative of a correlation with infection keeps a study grounded in what is really meaningful.

If you want my take it is the dynamics of markets which will decide this now.  Pfizer is established as the de-facto default for rich countries and has booked multi-billion unit orders in the US and EU.  It sells its jags at a good profit.   Needing regular boosters is a good thing from a business perspective if you can sell your product as being superior on efficacy.

Pfizer is going to have more revenue and it's going to be able to attract more investment.  This will mean it has larger scale manufacturing, a larger R&D budget and the ability to acquire competitors or suppliers with interesting tech and use money to attract key staff.  It will get more and more expensive to stay in the game.  The next phase is consolidation and a shake-out of suppliers.   I wouldn't be at all surprised if in in a year or two the group at Oxford have a new sponsor, Valneva has been acquired and AZ has exited.

I am having trouble with their website, but rising positivity doesn’t sit well with my LFD suggestion, so more towards opening nightclubs without enough infection or vaccine derived immunity vs england?  

I'd be interested to know what the 'major difference' is?

I think thread #38 has made some things painfully clear.

That's not super helpful!

Anyone else willing to tell me what makes Valneva so different and potentially so much better than existing vaccines? Wintertree seems to think it's a game changer which is very good news.

Basically the other vaccines are targeting just one area of the virus (the "spike" protein) to "train" our immune response.

The Valneva vaccine is using (I'm not totally clear on this) either multiple areas or the whole virus (deactivated) to "train" our immune response.

Which means that the virus would have to significantly mutate in several areas to become vaccine evading. Multiple mutations are of course much less likely than single mutations so this vaccine should theoretically provide better protection for the future.

Thanks Michael. I get that it's a Euro version of Sinovac/Sinopharm-type vaccines which are amongst the best established on the market.

I'm wondering what the major difference will be with the Valvena one and how it's going to make it better.

If you knew how it was different, then why ask?

I only know how it's the same as existing vaccines.

Edinburgh festival. 

Compare to what Boardmasters has caused in the South West.

Covid numbers in my brothers city in Vietnam. 

Thought you might be intrigued.He is under enforced lockdown in his area  

Very similar to Thailand, and Cambodia went a bit earlier - it seems delta is too much for multiple East Asian countries who had been successfully at controlling earlier variants?  Not good whatever the cause.  

Surely if they're only report PCR tests then if new cases are being found in Lateral flow tests the positivity, assusming this is only for PCR tests, would go up?

It seems to me that countries that previously managed excellent border control have left themselves very susceptible to Delta unless they have sufficient vaccination levels.

I can envisage a scenario where in the long run the UK ends up less affected overall because our poor policy and decision making has given enough of the population a level of immunity before Delta came along.

Whilst I would welcome that scenario for the UK long term, the thought that various politicians would be able to blithely go about claiming they got it right fills me with dread.

"It doesn't support your claim anywhere at all"

To avoid this becoming as bad as the 'ding-dong' with Tom, people can just look here and make up their own minds on who is right:

https://www.ukhillwalking.com/forums/off_belay/fncp_36_cont-737539?v=1#x949762...

It was never 'my claim', it was bruxist who picked up on a SAGE point partly buried in appendices, typifying our government's terrible communication on important public health matters (I'm guessing, in this case, because it clashes with their political problems with their own backbenchers on masks). Wintertree described the argument as 5d and bruxist acknowledges we can't prove it but its the best advice we have on the subject looking at real world use. Taking about 'Flimsy facepants' is simply a terrible scientific public health spirited approach.

Si dH suggested the Scottish data page excludes LFD results, so an increase in positivity would be unrelated to those switching on, unless most of them go for confirmatory PCRs.  

https://coronavirus.data.gov.uk/details/cases?areaType=nation&areaName=... says:

• Number of people with at least one positive COVID-19 test result, either lab-reported or rapid lateral flow test (England only), by specimen date. Positive rapid lateral flow test results can be confirmed with PCR tests taken within 72 hours. If the PCR test results are negative, these are not reported as cases. People tested positive more than once are only counted once, on the date of their first positive test. Data for the period ending 5 days before the date when the website was last updated with data for the selected area, highlighted in grey, are incomplete.

This also suggests LFD data is not reported for Scotland, unless it is PCR confirmed.  What I don't know is what fraction (if any) of positive LFD test results are sent for PCR confirmation in Scotland.  If most do go for PCR confirmation, this would raise positivity as LFD-gating into PCR tests should have far higher positivity and symptomatic-gating into PCR tests.  If most don't go for confirmation, it should have little effect.  So, who knows!  A real shame PHS can't/won't publish the PCR/uLFD/cLFD breakdown PHE do.

The numbers look exceptionally large to be LFD switch on with the most recent data point however, so I was probably barking up totally the wrong tree regardless.  Suggestions from Tom and LSRH: the opening of nightclubs and other venues and the fringe and kicking in (I had no idea that was on till LSRH mentioned it just upthread; I think that puts paid to another posters claim that "I think England has decided the epidemic is over, Scotland hasn't.  Scotland is pretty happy with Sturgeon's caution, England seems to be OK with Johnson's risk taking.").  

With any effects from schools ready to start feeding through in the next few days, it looks like incautious policy in Scotland is going to see cases charge ahead - not necessarily a major problem for healthcare if the growth is in the young (where lack of vaccination is concentrated) and doesn't force a lot of growth in older ages.

It's increasingly hard to get good meaning out of all this data, it not being categorised by vaccination status. 

A recurring suggestion has been that the generally lower history of infection in Scotland means there's more predominantly young people now lining up to get their first dose of antibodies au-naturale. 

The latest ONS antibody survey landed yesterday, it doesn't strongly support that:

https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/con...

• In England, it is estimated that over 9 in 10 adults, or 94.2% of the adult population (95% credible interval: 93.2% to 95.1%) would have tested positive for antibodies against SARS-CoV-2, the specific virus that causes coronavirus (COVID-19), on a blood test in the week beginning 26 July 2021, suggesting they had the infection in the past or have been vaccinated.
• In Scotland, it is estimated that over 9 in 10 adults, or 93.5% of the adult population (95% credible interval: 92.2% to 94.6%) would have tested positive for antibodies against SARS-CoV-2 on a blood test in the week beginning 26 July 2021, suggesting they had the infection in the past or have been vaccinated.

So, a surprisingly small difference.  

I think it's swings and roundabouts here - the population immunity we gained the hard way has probably blunted delta's spread compared to what we're seeing in parts of East Asia and Australia, but it came at a price of deaths, health damage and damage to healthcare. Meanwhile,  the knowledge about clinical care and the range of treatments available has come on so much since our first wave, that many of those who then died would now have lived, and likewise many of those who lived with serious consequences would have had those blunted.  The countries who held the line for the longest now face this with more knowledge, techniques and medications at their backs, and for Australia and now potentially NZ with at least the most vulnerable vaccinated.  It'll might easy for politicians to point at high cases numbers still exploding out of control in Australia in a months time, but it will not mean the same level of catastrophe and loss that such case numbers meant in the UK last March. 

I say "probably" because those who have managed the best border controls haven't had the same progress with vaccination - in some but not all cases down to the development and wealth levels of the nations - and it's hard to disentangle the two effects.

The sooner we can get cargo flights going out of Teeside and Edinburgh with the output form the Novavax and Valneva plants, the better.  All gone a bit quiet on the Novavax front other than the recent FT article on their delays going for certification in the US.   That vaccine in particular is robust against stage and shipping in less than ideal conditions I think.

I have had similar thoughts about our failures being misrepresented as some strategic masterstroke of success.  It makes me feel a bit sick TBH.  

I guess for a lot if the more cautious nations, working on the irradicatation strategy, it's about vaccine availability. Currently I'd still rather be in a country like NZ's position than ours but it could still go disasterously wrong, with only about 20% double jabbed.

https://www.health.govt.nz/our-work/diseases-and-conditions/covid-19-novel-...

Ahhh I had assumed that all positive LFT or at least all LFT reported, were sent for PCR. 

seems like even if this is a factor it's not the dominant one though it feels like if you're sending a proportion of people who are going to being 90% plus positive for PCR this would affect the positivity rate even if it doesn't have the biggest effect on the absolute numbers. 

At least some countries followed a consistent and honest approach doing the best they could for their population. Philip Hammond's new book is out as a collection of his Private Eye columns. Kindness is part of his key advice but that doesn't extend to our disaster of a government during this pandemic.

https://www.private-eye.co.uk/books/dr-hammonds-covid-casebook

Seperate thread here:

https://www.ukhillwalking.com/forums/off_belay/dr_philip_hammonds_new_covid_bo...

Is the following about right?

Most* people (vaccinated or unvaccinated) have not been infected.

Unvaccinated without antibodies heavily outnumber unvaccinated with antibodies.

Hence that variations in infections between countries not very significant compared to vaccination rates because immunity in UK comes overwhelmingly from vaccination rather than infection.

*from memory an estimated 15% of UK has been infected

"I'm sorry to see bad-tempered posts proliferating on this thread, with its pedigree of reasoned, polite and informative debate."

Well over a hundred thousand have died in the UK and we have been heavily restricted in our lives based on necessary public health grounds with consequencial negative health effects and the economic prospects of many badly damaged. It's an emotive subject with plenty to be angry about. The vast majority of content on these threads has indeed been incredibly useful and good tempered but not all of it was well reasoned or always polite.

I admit I am one who is less well tempered at times, but I usually don't regret that: we would be inhuman if we didn't become emotional sometimes in this disaster made much worse by disastrous government responses. I also think those of us who claim to be on the side of science need to be held to account at a higher standard and occasionally some pretty wrong things have been said here, like predictions of the 18 to 24 age group limiting at 50% vaccination. Wintertree sets an excellent example... analyse but don't predict. I have no time for pop-up accounts spilling antivax and covid denial shit but regulars making reasonable, if probably scientifically incorrect, based points should be treated better when their errors are pointed out, even when its clear the message isn't getting across. There are many examples where regulars haven't been treated very politely at all on these threads and I don't think that is anywhere near a moral high ground (and sometimes that includes me, which I do regret). 

Edinburgh Festival is tiny this year, perhaps 1%-5% of normal.

Have been to 5 events so far.

Book festival audiences - indoors, masks, socially distanced audiences of 26 (I counted) to 50 (I didn't count) rather than 200. Closer to one event at a time rather than 5-10 in parallel.

Fringe - socially distanced, masked indoors, capacity 20% of normal. There wasn't a queue for the next show when we exited! It doesn't have the usual organisational precision of a 60 minute show every 75 minutes.

International festival - audience 500 rather than a couple of thousand in Usher hall. Impressively engineered giant marquees, socially distanced but essentially outdoors (ie you need warm clothing) as there are no walls. 

Places like Pleasance courtyard with 10-15 venues, audiences ranging from 10-1000 with 10 shows per hour are closed. 

Fringe venue numbering goes up to 500+ but looking at Fringe website I think only a third are operating this year.

Although I think legally all venues can operate normally at full capacity it was all planned and set up long before that was possible.

I don't think hotels are charging the normal festival rates.

I only got offered one leaflet on the Royal Mile!

There are visitors (like us) but it may not be very significant compared to pubs, nightclubs, cafes & restaurants in Edinburgh or any other city. 

I think the rise in cases has arrived just a little too late to be explained only as an Edinburgh festival effect (Festival started 6th Aug), and furthermore the rise in cases is fairly co-ordinated and widespread across Scotland and certainly not most obvious in Edinburgh. It is noticeable that East Lothian (right next door to Edinburgh), which had very high case loads back in June, has not yet had an increase in cases. Nor Dundee, actually. So I don't think this is explained by an Edinburgh effect - if anything it looks a bit more Glaswegian.

It's all speculation really.

It'd be nice if the immunology partners in the ONS antibody survey could run a really granular analysis on the kind(s) of antibodies as there are measurably different responses for the vaccine and for infection, although there's probably ambiguity over the combination of "infection in wave 1 + vaccination" as they share a common spike protein.  Then again, many of those people will have reducing serum levels as the information shunts over to immune memory cells.

The ONS readily disclaim that reality is a lot more complex than a binary [yes/no] to the question "got antibodies?".  How precisely one produces a more meaningful metric - I suspect there's so many variables around immunity and around infection that it could lead a lot of smart people on a merry chase for months and not do much to improve understanding.

The most useful thing now would be for all the dashboard data to be broken down by vaccination status.  The data must be there to do that, and at least one dashboard now does this internationally (Israel) and I'm seeing a lot of data like this out of San Diego county.  Good, hard data that can be put together more easily than a statistical understanding of the kinds and levels of antibodies have, and data that is a direct readout of the situation unlike the antibody data.

Mind you, I shudder to imagine the CSV or JSON monstrosity this would manifest as on the dashboard - they don't seem to be a great believer in regularised gridded data formats, so the UTLA + age breakdown has a single line-item for every age bin for every UTLA, with one row looking like this -"E06000014,York,utla,2021-05-23,75_79,0,0,0.0" - and something like 1.5 million rows.  I keep avoiding a grumpy old man rant over spectacularly bad use of data formats...  Served uncompressed as well...  Probably explains why the dashboard API server is hammered to the point of instability for the first hour after each day's data release...

Hi Offwidth, I replied to your post because it was the one at the bottom of the thread – it wasn’t that I was singling you out or anything. And I agree, we all have good reasons to be angry sometimes – even when they aren’t obvious to others. But if conversation can be conducted politely it is generally easier to discover and build on common ground – it’s so much more interesting and productive than entrenching positions and lobbing shells. I think wintertree's plotting threads have in general had a greater degree of sharing and building than is common on UKC. I like it, that's all.

On the charlatans and pop-ups, I am of the school that thinks they are probably better ignored – though it’s easier said than done.

I don’t know why it’s getting so fighty around here; normally it happens when everyone is locked down.  A lot of bad will was generated quite broadly from the recent threads severely misrepresenting the YCS data I think.

I’ve been thinking about starting a separate “Friday night science fight” thread.  I’ve got a corker to get it started…. 

I have looked at some API data but couldn't face parsing it!

So, a surprisingly small difference.

Yes - and given subsequent vaccination rates - likely to have got a little smaller since. I am getting a strong sense of having absolutely no idea what's going on again.

I was pretty sure its wasn't just about me but I think the points I'm making are important. 

Science, as a clearly defined method of understanding aspects of the world around us, is not the same as the outputs of the 'industry of science' including government, industry, the NHS, independent research institutions and academia (an output which is normally in the UK just about OK and is too often shameful), nor always are the views of individual scientists claiming to be scientific anything to do with science. As such we need to be especially careful with those presenting a scientific line as I think it is important to call out badly presented information and especially so for any bs.

As an example, the football effect on covid case growth has good evidence now but many other areas blamed here are just a guess, not evidenced anywhere and not worded appropriately given that, if coming from someone claiming a scientific approach.

A current example of covid specifics in the 'science industry' that really bug me and I think should worry anyone who believes in openness in science (and made me grumpier than usual this week),  but barely seems to have touched national news, is the Roy Lilley exposure and campaigning this week on the ridiculous  tight and undemocratic control of NHS trust communications to avoid scientific fact embarrassing the NHS leadership and government.

I think any pop-up covid denier and antivax accounts need exposing publicly as such, alongside a report to the moderators, and a simple polite position on why they are wrong and be ignored thereafter as they will soon be gone.

https://www.tandfonline.com/doi/full/10.1080/23744731.2021.1944665

Knit cotton mask, i.e. 'shabby facepants', 15% reduction in aerosol conditional infection probability. Properly fitted mask with clip on the nose, much, much better than that. Not going in there, 100%. Your repeated proclamation that face coverings are the best defence, without all the necessary caveats, is downright dangerous.

Back to SNAFU.  I like the rare moments we can piece together enough to be confident we understand what's going on.  Normally it happens 1-2 weeks after it happened, so sudden change => sudden discomfort.

Update plots below.  The rise is still landing in Scotland.  I don't know where Travelling Tabbby gets their demographic data from but I did a quick ratio of 'recent low water mark' to 'current level' for cases/100k reading off from his plots; it's a bit deceptive as more of the growth has perhaps happened more recently in the younger ages, meaning how rapid the growth is is more skewed to younger ages than this suggest.  

00-14: ~1.7x
15-19: ~2.2x
20-24: ~2.8x
25-44: ~1.5x
45-64: ~1.6x

This is the growth ratios but the absolute numbers are highest in 15-19 and 20-24, so the absolute growth their is doubly big.

So - not quite panic stations (easy to say living comfortably south of the border), as this growth is demographically concentrated in young adults so shouldn't lead to an equivalent growth in hospitalisations (by a long stretch).  Without longitudinal data on vaccine status its hard to interpret if this is the opening of nightclubs (&c.) leading to a rapid fill-in of the gaps in immunity or if it's something else...

To put it in to context, the growth in cases from the previous reasonably steady point is ~1000/day, and there's perhaps order of 200,000 people without antibodies...

The doubling times now (may not be the worst to be seen, still getting faster) are compatible with those seen during the football so not "stand out" when it comes to mass indoor socialising.  The question for me - riding on the massive assumption that this is the last round of Scottish unlocking...

• Why did opening night clubs cause this spike in Scotland but not England?

It would take some set of coincidences for the 3-week difference in the shutdown of the football effect to have boosted immunity in those affected by the last unlocking enough to make such a big difference between Scotland and England.

In other news, the first Covid MAB has been approved by the MHRA today. 

https://www.sciencemediacentre.org/expert-reaction-to-mhra-approval-of-rona...

I read that nightclubs in England were struggling a bit, cancelling nights as demand wasn't all that great, certainly nothing like normal. 

Today have read 9/10 still wearing masks (although maybe not quite as much as before?) and 9000 cases linked to euros https://www.theguardian.com/world/2021/aug/20/9000-covid-cases-linked-to-eu...

Why any of that would be different in Scotland, I don't know though, other than the football. Maybe our nightclub infections got lost in the football ones? 

Of course 'not going in there' is the best.

But we're trying to move out of a complete lockdown. Assuming we want to do that, and return to some level of normality where 'going in there' is allowed/desired/necessary, then wearing an effective mask is probably the second most effective precaution, after vaccination. I don't think Offwidth claimed masks were better than vaccination, or that they were the 'best'.

Bit stumped by the Scottish thing. Your graphs make it look like we might still be seeing residual effects of level 0 day amplified by the final openings inc nightclubs, rather than the latter in isolation. The rate constant has been increasing continuously since approx 27/07? (Edit- I can see it did behave differently for a few days in the first week of August actually.)

Their absolute case numbers have been lower than England recently. Maybe they will start to flatten off at a similar level (not there yet) now that restrictions have been removed in both nations, if immunity levels are similar. Speculation.

Re: the new antibody treatment...it sounds great in principle, but apparently it's expensive and it doesn't seem clear how/when it would be used, given the indication that immediate treatment is best. Seems like something that would be best targeted at anyone over 65 immediately on them receiving a positive test result, if case rates were low enough for that to be affordable. I suppose you could limit this sort of approach instead to people who were both old and unvaccinated, if necessary.

Thanks for that link.  

It's really interesting to read the summary findings on the numbers and capacity fractions for other sports vs the football contrasted against the amount of infection directly spread at the events.  A time for the institutions and media channels profiting from football to step back and reflect on the monster they have created.

A good thought.  I get the impression in England that there was a real flurry of activity when they re-opened that hasn't been sustained; if that did have a brief duration spike in transmission it could have been masked by the post-football collapse; England was a lot more invested in the football by the time we bombed out, but the change in rate constant was similar.  Trying to quantify anything disappears in to a range of intangibles; but if my "flurry" comment and your "buried in football" theory are both on point, we can deduce that the rapid growth in England was't sustained, so we'd expect it to collapse back down in Scotland in a few days - except that the switch-on of schools will taper in as that effect would taper out, but the demographics will change.  I found the link to the PHS source on the demographic data for Scotland so I'll try and do a Scottish demographic rate constants plot for tomorrow which might help test some of these ideas out.

In reply to Offwidth:

Looking at the SAGE documents I often wonder if there's too much investment in speculative modelling and too little in high quality analysis and presentation of nowcasts on the data.  It's easy for scientists to look at a few plots and grok what it all means, but in exponential situations that's far from guaranteed for people from broader backgrounds, and as we've seen (from people at respected universities down to pop-up trolls) it's easy to misunderstand or misrepresent the data as a result.  A really high quality data analysis and presentation unit would be a wonderful bridge between science (done by scientists who tend to forget the decades of experience they're drawing on when interpreting and talking about data) and the other 98% of people.

To my reading a lot of the discussion here is speculative - putting qualitative ideas out there on what might be to blame, and then there're attempts to see if the data does or does not support that (which is not the same as it being true or false).  There are sometimes some egregious statements without support which tend to get jumped on.  I know I've suggested a few things off intuition and had my feet held to the coals to try and evidence it.  There's a worry people reading the thread might take suggestions as gospel but I hope not, I tend to assume the people reading this can spot the difference between data driven discussion and speculation.  I have found the speculation endlessly useful in terms of holding it up to the data and seeing what sticks.

Indeed.  I had a meal indoors last week - 9 of us booked out a 30 person capacity private room and sat far apart.  Bit disappointing in the end; only restaurant within 15 miles to serve Ribeye and I could have cooked it better for 1/4 of the price.  Still, I didn't have to do the washing up...

I agree with the sentiment here, but I think proper air filtration or turnover is more powerful than low grade masks and I see far too little focus on air handling(*).  For delta, in a moderately crowded indoor location it looks like you want both.

But all these measures are multiplicative - if ventilation halves your risk, masks halve your risk and vaccination halves your risk, put them together and you've got 1/8th of your risk.

It all comes down to the messaging about how each measure helps.  I feel it's pretty intuitive to people that the more measures you take the better, and that one doesn't obviate the need for another.

I don't think I've seen an argument go on at cross purposes for so long with so much frustration before...

(*) re: air handling - Except my dentists - we had a great natter about infection control whilst we waited for the anaesthetic to take hold the other day - warmed the cockles of my heart to see the forced airflow setup and the HEPA unit in the obvious dead air. 

You're not the only one stumped by Scotland.

Cases rocketing since July 31st but not feeding though to hospitalisations and deaths which are still declining 20 days later. It's great news, but puzzling.

Cases are concentrated in younger ages but those people have parents, grand parents and older colleagues/customers/friends/whatever. I wonder how long that takes for infections to diffuse between different age groups.

I may have had a few glasses of wine, so forgive me any lack of clarity.

This is exactly what's been unnerving me. It's hard to think of any other explanation for the rise in Scotland - and yet in England the equivalent (supposed) cause had no apparent effect. Conclusion: I am in the dark.

SidH and Kirsten (in their posts just above, thank you both) independently offer a new thought for me - that the Scottish rise is perhaps just a crease that has been ironed out differently in England. It's an obvious point really, but one I've failed to see previously because of my focus on catastrophic growth and the feeling that, without action, it is inexorable. I suspect I have not alone in that set mind. But, post mass vaccinations, we're in a new territory - as you pointed out some while back.

I made some vague observations and posed a vague question, and got some clarity and good ideas in return - UKC doing its best.

That all seems reasonable as speculation goes.  Two lowerings of restrictions, two up ticks in rate constant - some masking from the collapsing football and a minor fakeout from the weather.  Hopefully somewhere behind closed doors they’re rapidly deciding if nightclubs were a step too far…

Re: the MABs and cost, almost anything is cheaper than time in ITU.  Taking upper end cost estimates for the compound of £2k for a dose, that’s around five days in hospital or one in intensive care; so plenty of scope for it to pay for itself - given the right risk and benefit profiles.

Good stuff. I'll start with a further point on ventilation and politics from Roy Lilley's review of  Dr Hammond's new book on the covid crisis:

"As Ian Hislop says;

‘The book is better than the Public Inquiry - and quicker and cheaper.’

In March, Hammond wrote about ventilation and how it was a priority for the US Centres for Disease Control and Prevention, urging schools to use portable air cleaners and ventilation. Meantime, in the UK; ‘…filtered flowing air is not seen as a priority. The typical state-school classroom contains 31 people, has poor ventilation, and teaching periods of up to two hours before teachers and children leave for a break…’ In a direct challenge he lists eight statements that go to the heart of HMG’s press blah, blah…

We were well prepared for the pandemic

Herd immunity was never the plan

We put a protective ring around care homes

There were no PPE shortages

We protected the NHS

Our T&T system is world class

Our border controls are world class

Our road map our of the pandemic is irreversible 

He says; anyone who makes any of these statements is a fool or a liar. That’s communications for you."

These were all what I would regard as output of the 'science industry' of government and all dishonest which shows a clear public failure of the government scientific establishment to hold the politicians to an honest position in a national emergency. Some failure is inevitable in the nature of politicians but the scale and consistency of this communication disaster should have been dealt with by the scientific establishment.

Back to masks. In your good explanation of multiplication of small gains explanation, everyone seems to forget our bodies defences. If the high viral load larger droplets are blocked or the aerosol 'fug' is reduced, the difference might mean a particular body might be able to fight off its exposure to the virus without infection. Not everyone gets infected where there is an indoor outbreak, and small differences might be enough to help increase that number. Becoming infected  is unlikely to be a linear process vs viral dose, so It might be better than multiplication in the 5d complexity of infection. Indoor, public-health mask policy is about protecting others from an infected mask wearer. The USS Theodore Roosevelt isn't a proper controlled experiment but it does indicate wearing a mask likely offers some protection to the wearer. This could be due to the mask or some other causal factor like correlation to those who will behave safer, something which needs investigation to determine.

I see speculation as useful  but if someone claiming a scientific position speculates they need to be clear on that and careful. That person should behave better as well in dealing with ordinary people who make scientific mistakes and reserve any ire for genuine bad actors. Exaggeration for comic effect and satire are great tools in exposing charlatans but in scientific debate it is at best a sign of insecurity. Scientists should be held to account at a higher level on science and apologise clearly when they get things wrong.

If you look at the cases heatmap for England (on the Covid dashboard) you could perhaps get some idea.  For the beginning of the 2nd wave, and the current wave, prevalence increased rapidly initially in younger people, but then took about 3-4 weeks for a peak in older adults

I very strongly agree with this sentence.

I think seeking out and citing whatever quantitative numbers and peer reviewed evidence exists is the right thing to do, and speculation should be retracted, or at least curtailed, when contradicted by data. And any statements presented as fact should be verifiable.

An unusually clear and IMO accurate presentation of the questions around “normalising” this virus into circulation with all the others.

https://www.bbc.co.uk/news/health-58270098

I thought the comments on “over vaccinating” whilst hypothetically sensible were in reality unjustified, pre-empting as they do the JCVI guidance.  My expectation there is that the criteria will be around a medical risk / benefit analysis rather than a blanket “boost everyone” approach. 

That's a good read with decent straightforward explanation (for those of us who don't understand the more complicated details of virology!), thanks for sharing  

I'm not so sure that everything said in that article is accurate. Manaus faced overwhelmed hospitals and huge levels of mortality twice, so I'm not convinced infection based immunity is anything like as effective as immunation based immunity, and reinfection data shows herd immunity through infection is clearly a highly dubious idea with the added higher risk of new mutations as countries 'run hot'.

I do agree boosters might be of marginal benefit for most double jabbed healthy people in western nations but are vital for the health and economic prospects of poorer countries. We can't live a normal life with open borders in the world until we are on top of hospitalisation in the vast majority of countries and I fear some of the poorest countries might end up isolated for a long time.

Re: Manaus, herd immunity etc.  My take:

There’s a difference between immunity granted to an adult by “virgin” (without antibodies) infection and that granted by vaccination then infection.  

Understanding that difference is a key step.

Herd immunity is almost certainly off the table now by any means; the virus won’t be eliminated in the UK let alone globally by immunity alone.  It’s not about achieving herd immunity but achieving a situation where it can circulate without causing any more harm than many other widely circulating viruses.  Still something of an article of faith that this will be achievable - but we won’t know if we don’t try.

Well, don't just stand there boasting about your "corker". Come and 'ave a go if you think yer hard enough...

More seriously, though, I don't think I'm a fan of all this virtual, socially-distanced fighting. I find it tends to make me blank out whilst reading through large sections of threads, as if I've found myself unwillingly eavesdropping on a private argument, and I'm sure I miss useful stuff because of it.

It‘a delayed until Sunday because I need to re-visit some old data I put together on the landscape; it’s well out of date…

Here’s to an argument free plotting #40.  Spoiler alert - I’m glad it wasn’t me who pitching the Scottish approach as more cautious than the English one a few days ago… 

I think there's two different aims & measures getting mixed up in this debate.

For getting 'back to normality' at the population level, vaccination is definitely superior to masking.

For preventing individual transmission events, masking (of a specific level, and for protection rather than source control) is definitely superior to vaccination.

Scotland very obviously is more cautious.  We went slower on the removal of restrictions and we still have mask rules.  I call that more cautious.  Scotland has been more cautious throughout the pandemic in terms of when lockdowns were started and ended.

I do have the feeling that you both, Offwidth and Longsufferingropeholder, are arguing at cross-purposes, and are where it really matters in broad agreement with each other - though this might come as a surprise to you!

There are two contexts neither of you has mentioned, and I suspect each is arguing from the point of view of a different context. The first context is time.

If the question is, did cloth face coverings help mitigation in the past, then I suspect you both would consider that they did, though perhaps to a different degree. When the dominant strain was alpha, ventilation was a thing that we here on UKC, espcially Wintertree, were banging on about but was nowhere to be noticed in Gov guidance, which was still heavily focused on fomite transmission. In this situation - that of a less-transmissible strain during a period when airborne transmission had still not been officially acknowledged and Gov guidance was mostly misguiding towards mitigation measures based on flu - it seems to me that face coverings of any quality were almost the only covid-specific method of source control that we had. There's one caveat to that but I'll come back to it later (1).

If the question is, on the other hand, do face coverings help mitigation to such an extent now, I suspect you both would be agree that things have changed. Delta is much more transmissible than Alpha; we now acknowledge airborne transmission; a close contact has been redefined as being within 1 metre of an infected person for 1 minute. To my mind, the effect is that source control mitigations are weakened, whether they're masks or any other type, and protective measures are preferable. I for one look enviously at Germany, where FFP2 masks were mandated and provided for free, and where waves of infection map onto the UK's but at a fraction of the prevalence.

The second context is the purpose of mitigation measures. Offwidth is arguing from the point of view of someone whose main concern is to reduce the risk of transmission from someone who unknowingly has Covid but is unavoidably in an enclosed space with other people. Longsufferingropeholder is arguing from the point of view of someone whose main concern is to reduce the mixing of the infectious with the as-yet-uninfected. These are two different purposes requiring different measures, as the original SAGE docs outline.

N.b. if I have misrepresented or otherwise misunderstood either O's or LSRH's views I apologize and am confident they'll correct me instanter...

(1) Back to my caveat about the UK emphasis on home-made face-coverings last year. My feeling is that UK Gov PPE procurement failures forced Gov into directing the public away from effective masks and towards the homespun. The result was a public that genuinely thought they were doing the best thing possible by knitting their own face-coverings, as if this were the 1940s. We're left with this Etsy cottage-industry notion of masking now, and whilst it might have helped in all sorts of ways against alpha its effectiveness is drastically reduced against delta - hence a large number of the UK public sees masks *as a whole and of all types* as pointless. That's a tremendous pity, as what we would once have regarded as high-grade PPE is now amply available, and we know it has the potential to inhibit transmission to a much greater degree than the current vaccines. But then, UK Gov strategy now seems uninterested in inhibiting transmission, and it's difficult to judge whether or not that's the right call as there's no clear statement of what the aim is.

My concern is both, but it's hard for part of it not to get lost along the chain of replies. I think we do agree about everything but the priorities and the messaging. I think that's all we differ on but it's easy to lose that in the multiplayer notepad debate format. I'm really very anti anyone saying things along the lines of "it's ok to do x if you wear a mask". I've never denied or doubted that they reduce the risk (1), but that's all they do, and not by as much as we'd all like to think they do. Far less in fact, unless really very carefully fitted. That's my one and only beef. My position is that you shouldn't be putting one on until you've asked yourself if you wouldn't be better to go elsewhere. It's the "unavoidably" that I'm chipping away at.

(1) both of giving and getting, more so of giving. Again, no departure from popular wisdom here. 

Hope that clarifies and doesn't perpetuate.

I hope we aren't going to argue (because I will inevitably lose🙄), and it wasn't me who said the Scottish approach was more cautious..... but I don't think you can reasonably deny that in comparison to England the Scottish approach has been more cautious, can you? Compared to England, restrictions were held on to a little longer in the summer of 2020 and re-imposed a little more rapidly as the second wave started breaking, particularly around Xmas. Scottish pupils were not, let's remind ourselves, sent back to school immediately after Xmas for a single day.

More recently England's bold bacchanal of Freedom Day was enacted as a more sober two stage affair in Scotland, and in addition we still retain mandatory mask wearing on public transport and public indoor spaces. Perhaps more important than all of that, the messaging in Scotland has all been about taking it seriously and protecting each other; not about shaking Coviddy hands and, ironically, Freeeeeedom! In any normal sense of the term, I think it is fair to say that the Scottish Government has indeed been more cautious than Westminster.

And to prove the pudding by eating it? The death rate from Covid here is approx 2/3 that in England. (I admit this is a cheap shot, because we don't know to what extent, if any, this is related to the cautiousness. And by the same token, we can't claim the rise in Scottish cases in the last week is on account of any incaution.).

Full disclosure: I detest nationalism, even the relatively benign nationalism of the SNP. If I prefer tha actions of the Scottish government over those of the Westminster government it is not because one is "Scottish" and the other "English".

I agree that on a like-for-like comparison the measure in Scotland have been consistently more cautious than those in England.  But...

Indeed - and so infection granted immunity is lower - especially with the relative differences in the switch-off of the football effect - and so this raises the possibility that "cautiousness" is not a directly comparable measure, but one defined locally in terms of the local situation.

This is the difficult situation Australia and NZ are in - with almost no infection granted immunity (which will tend to land where it's most needed / beneficial) they have to be far more cautious until vaccination is complete.  As discussed up thread I absolutely think they did the right thing to protect their people, and whatever happens now, net total it will be less bad than if they'd lost control 18 months ago.  

It's something of a paradox that the more cautious somewhere has been, the more careful they have to remain until vaccination is substantially complete - and the more gradual they have to be about allowing the natural circulation that must inevitably follow, flashing through the gaps in immunity light lightning.

So I think we're both right and we're both wrong....?  

But if I was the Scottish Government I'd be pulling the cliched Hollywood "Shut it down, shut it all down" act over nightclubs tonight until the epidemiology is done on the recent rise.

Excellent post.

Exactly this. And once that precedent was set, it was impossible for this government to admit that their extemporaneous PPE advice was flawed, and that more effective masks should be used.

"It's okay" is not, and has never been a binary assessment.

All measures only reduce the risk. None eliminate it. As wintertree says, measures are mostly multiplicative (to whatever extent, and provided they don't duplicate each other).

So, whilst I'm fully vaccinated, and mostly keeping to myself, I still wear a mask in shops and on my rare uses of public transport.

I am taking preventive measures to reduce, not eliminate, the risk of me spreading the virus unknowingly, whilst slowly returning to more normal life.

Wearing a mask is a simple, cheap measure. Improving airflow and installing HEPA filtering is more expensive and not so simple, even if it might be more effective.

The way they've been acting up to now my guess is they probably will notch it back up a little because of the latest numbers and nightclubs are the obvious thing to take action on.   IMHO it was a mistake to let nightclubs open until after the age group that uses them had all had their second dose and a few weeks for it to take effect.  On the other hand I never go near them anyway so shutting them costs me nothing.

So this does beg the question (and this genuinely is a question), was the original scientic advice to let it spread a more reasonable decision than it first seemed? I appreciate we are looking at a snap shot in time right now, but the UK appears to be in a relatively strong position.

I realise that a lot of things have happened since the initial decision making (vaccines/variants) that substantially alter the position we are in.

I thinks it's clear that locking down to late on numerous occasions was crackers, but was the plan to allow it to spread more reasonable than maybe it first seemed?

Just to reiterate, before I get shot down in flames as a governement apologist, I'm not. I'm interested in the views of the hive mind about whether this was pure luck or a bit of judgement.

I see the question.  

In my view, given mass vaccination, no.  The vaccine is the larger contributor - by far - to why the UK can run cases so hot with “only” 700 people/day going to hospital.  (Far from a small number) and “only” 70 deaths/day (in the summer).

Critical point: A lot of the infection granted immunity likely contributing now is from an era where it spread with those more vulnerable to harm protected by near 100% vaccine uptake, which is totally different to when spread was being urged last March by some.

To have allowed it to spread enough to get us to this point without vaccines would have meant over half a million extra dead in six months.  Without vaccines we’d have had to try elimination or try the “megadeath approach”.  With vaccines an early rush to spread it is pointless death - but allowing it to spread after widespread vaccination of the vulnerable - and allowing spread through the vaccinated - becomes part of the solution.

The main problem for Australia and NZ is slow progress vaccinating.  They’ve not been able to have a wave of transmission of delta with significant vaccine protection of older adults unlike the UK.  

I'm inclined to agree, but....

 The vaccine is the larger contributor - by far - to why the UK can run cases so hot with “only” 700 people/day going to hospital.  (Far from a small number) and “only” 70 deaths/day (in the summer).

Totally agree

There was a big blind spot with respect to a few things. We didn't know when we would get a vaccine. We didn't know how effective it would be. And we didn't know what the future strains might look like. 

So option 1, elimination. This requires the rest of the world to do it. Or it requires a level of border control that Australia has not managed (and it could be argued that they are in a better position to achieve this)

We could argue about whether we could have achieved this, but one would have to accept that there's a reasonable chance in this less than perfect world that this may not be achievable.

If for sake of argument we assume that elimation is very difficult or possibly impossible, and we take a middle position on vaccine availability and efficacy (i.e. Far later and less effective than we managed), and we are concerned that the virus may become more dangerous , are we better off letting spread (I don't mean let it rip).

I'm clear the implementation of the policy was dreadful, but I'm not clear that the policy was totally unreasonable.

And if a vaccine wasn't available, which I don't think (I'm prepared to be corrected on this) was outside the bounds of what was expected, then the problem for Aus/NZ is that they would have a population twice as vulnerable as ours.

Until recently I couldn't see an argument for the policy. More recently, I see an argument, but it's not one I'm convinced of.

It's not the policy that's unreasonable - it's more or less what we are doing now after all - but the timing of it.  

If there was no vaccine, if elimination was off the table then controlled spread would be the only real option but given the maximum level of ill that can be accommodated in hospitals and the fading of natural immunity, I don't see how we could actually have achieved this with operational healthcare (without it its way more than 0.5m dead) in a sustainable way.

"No vaccine" is a disastrous scenario - It would probably need an approach including live viral inoculation of the lowest risk 1/3rd of society under their own isolation conditions with managed infections. Not "focused protection" so much as a division of society and speeding the process up where safe to do so, and slowing it down as much as possible elsewhere to preserve healthcare capacity, then once the "live inoculation" is done, the fraction of immunity reduces R making it easier to flatten the curve.  Net result if it works is still half a million dead, but without scenes like those seen in India or Brazil.  Unthinkably awful however it's manned - and what could have happened if this had come out of Asia 20 years ago instead of original SARS.  Shows how catastrophically unprepared we were 20 years ago.

Under a "no vaccine" scenario there are still massive benefits to controlling cases through NPIs to begin with - buy time to study and learn about the enemy

• We did not have the testing capability at the time to even try to manage controlling the rate of spread - we either slammed the doors down on growth or we had a catastrophe.  
• Waiting brought a raft of improvements to clinical care perhaps halving the fatality rate - and many more interventions are making their way through the trials pipeline.  

I've said before - assuming the cross immunity from original SARS and MERS is bidirectional, this virus could close the door to a future outbreak of another SARS like virus with their horrifying fatality rates in young adults.  There's shared viral proteins across the whole family, and new viruses from it keep appearing, so far 2 out of 3 are highly lethal.  I wouldn't want to bet against the next one to emerge.

It was't just vaccines - testing capability, re-organisation of WFH and food supply, lessons in clinical care, repurposing trials and new drug trials - none of these things were ready when it was proposed to push through it last March, and the masses of work done on each of these all contribute to why things are much less bad now.  It was simply madness to rush blind in to a burning building. 

This is only viable for a something that is effectively a totalitarian state. The only way it works is to hide from the population, until it becomes obvious, that of the order of half a million people in the UK are going to die. A few months in became obvious that another similar sized group would be left with such severe permanent lung damage they might wish themselves dead (an early-ish Dutch study showed everyone who had been on a ventilator ended up with permanent lung damage...a large majority on ventilators survived). Then later on it would become obvious that there would maybe be a similar number again so severely incapacitated with post-viral long covid that it made any physical exercise that relates to living a normal life impossible for over a year.

We would also have followed almost the same path, as in a herd immunity approach you still have to protect hospitals.

There is a final solution... don't hospitalise those with severe covid, so hospitals will not become overwhelmed. Deaths would increase by somewhere from a factor of 3 to 10 depending on how well we could devolve medical self help.

The Swedish, Dutch and British looking at this herd immunity approach early on were using the wrong modelling, getting too little advice from those who had experience in outbreaks,  and so were underestimating the virus and maybe pessimistic about control and vaccine chances. On the virus side, variants have made things a lot harder still. On the plus side, like often in  history,  scientific advance accelerated in a crisis, so vaccines arrived earlier and with greater efficacy than expected.

I don't get why some people think Australia and NZ (Singapore, S Korea etc) are certain to be doomed with Delta. Their plan has worked so far. They have effective  border control and TTT. I think the odds are still on their side.

I think my fencing partner is playing games. One minute  he's serious and the next he exaggerates, sometimes sarcastically. This is a forum so he can do what he wants (within the rules) but he can't then claim he is being careful about what he says and get taken seriously if he lectures others.  I've always accepted the realistic data on masks but the majority of people I saw in public wore reasonable quality basic masks well and socially distanced better than predicted (anyone else remember the dire mask predictions: people would go mad thinking they were safe?). I'd expect low mid percent protection from such a wearer and so, for anyone worried in shops say, being able to remain at pretty low risk given the combination of time and seperation. People wearing shit masks badly can be seen and avoided. The exception for masks is work and public transport where time and seperation are not on our side.

The homespun boom was useful in my view as it was a good intermediate solution in a PPE shortage and getting people involved reinforces the public health message: that will always help in a pandemic. It's probably still good enough for shops. That the government didn't change mask advice where it was needed was nothing to do with going down a blind alley. As PPE pressure declined the advice for better masks, especially in work, could easily have been changed in my view (who would object?), just like the advice for better ventilation, ensuring symptoms were still the right ones* and many other public health mistakes.

My impression from watching this catastrophe is viral loads vary a lot, from the infected but almost uninfectious, to much rarer super-spreaders. Infection is non linear...large dose droplets coughed into someones face is a lot more serious a risk than aerosol... some people with no previous C19 immune response fight it off next to people in the same room who become infected.  Even small differences might be enough to avoid an infection and.mask differences multiply in a non linear infection process. It's a major public health benefit easily done, not a magic bullet of immunity.  It's not unreasonable from the data to see most spread being at home and much of that outside the home from where precautions were being ignored.

Yes time has changed things but I'm not at all convinced R0 for Delta is 8, as the worldwide impact should be much larger. We are likely looking at magnified super-spreader events in a perfect petri dish of nightclubs with more covid denier and antivax sentiment than in the public. Yet what happed in the Netherlands didn't happen here. Still it doesn't matter for those who support the public health message: adults should meet as much as they can outdoors, get vaccinated, wear a mask indoors in public, insist on the best ventilation possible in indoor public space, social distance in public, and maintain good hand hygene, in something like that order

* I know someone double jabbed, NHS employed, who suspected they might have covid from current symptoms they had heard for Delta, but not matching the official symptoms, and felt they had to lie on the telephone triage to get a PCR booked. They were positive.

This. 100% this. If you'd always said all of this, I'd be totally with you. I'd swap 3 and 4, but if they're given at least equal weight then we've come around to violently agreeing with each other.

Quite. The same logic has cropped up countless times before with every H&S intervention, never with any justification and always proved wrong in retrospect. In this case it's difficult to figure out where such ideas are coming from - behavioural science would be the usual suspect, but it's certainly not the behavioural scientists on SAGE, e.g. Stephen Reicher; more likely Behavioural Insights and the so-called Covid Recovery Group.

Absolutely. Yet a large number of people are seemingly unaware that this is still the message: if it's only a recommendation or an expectation, it doesn't count. Here's another area where (presuming these messages are being issued in good faith) Gov's behavioural science advice seems oddly ignorant of British proclivities.

I've just come to the end of two months based at local hospitals - am moonlighting away from academia in Public Health for the LA - and this week went back to the office. The morning I returned, my colleagues were participating in a 'well-being' event, meant to be some sort of post-pandemic spirit-builder, all indoors in one office room, mostly unmasked, with a buffet. I dodged the event. The following day one of them tested positive. They're all double-jabbed. It'll be interesting to see what happens over the next ten days. I'm supposed to do a daily LFT test, but access to PCR now includes exposure, so I'll be substituting an LFT with a PCR on days 5 and 8, fortunately without having to lie in order to get access.

That's my point and why you annoy me so much.

You are really funny and insightful at times. Even 'facepants' was funny if you say it just pisses you off how some people wear masks but instead you use it to repeatedly undermine the public health mask message and dishonestly represent people who support that message. I was just repeating a line from SAGE, via bruxist (shamefully buried in pdf appendices).

You also come out with childish crap like Indie SAGE is the Guardian. As a Guardian regular some of the opinions and leaders drive me spare but on their serious covid reporting they are the best of all the UK newspapers I know (and are free online). They like Indie SAGE have made mistakes but so has everyone.

You were also totally wrong on 18 to 24 age group vaccination rate saturating at around 50%, with no apology. In that argument I made a mistake looking at some data and apologised. I was still proved right overall... remind me of current numbers (and they are still based on flawed population data, so clear underestimates)?

I'd accept your order of public health actions and always would have,  SAGE gave the marginal preference to masks in the real world situation. They are pretty similar effectiveness from their data. Instead, you implied I was being irresponsible,  claiming magical powers for masks etc.

I'd swap 3 and 4 too, in ideal circumstances, but feel the UK doesn't really offer those. Over the last 8 weeks I've been unavoidably in hospital buildings built in the early 19th C, late 19th C, 1930s, 1980s, and one built in 2005. The mask issue hasn't really come up as masks are still mandatory for everyone in hospital - as an aside it's been interesting to observe that absolutely no-one claims exemption - but the difference in ventilation capacity is wild. If I were working anywhere built in the Victorian era - say, for example, the Palace of Westminster - I'd definitely prioritize masks in view of the best ventilation possible in that specific building still being less than desirable.

If I had to work in the palace of Westminster I'd be dressed for porton down

Heh! Your best bet in that place would be to wear a black hoody, thereby triggering all the worst MPs' ingrained social-distancing instincts...

I've a lot of time for your contributions too.

I always tried to be really clear that people should wear them at all times if genuinely they couldn't avoid those situations. I thought I'd gone to a lot of trouble to make that front and centre. I absolutely did not mean to encourage anyone to toss them aside, but to be aware of the relative magnitude of the protection they confer, and to give it some thought.

This is a separate issue that stems from my annoyance at the UK press as a whole. It manifests as anti-guardian here only because of the pro-guardian hegemony that pervades these pages. We should probably agree to disagree on the guardian. Or maybe I'll agree to read the guardian from time to time if the UKC hive reciprocates by spreading its wings a little. I've learned I'm on my own there so I'm not going to dig that hole any deeper.

I don't remember much of this battle. I do apologise if I was wrong, but I was following the various analyses, tracking the delivery numbers, and doing a lot of digging on that, so I'll be surprised if it's not just a difference of interpretation. Agree the population data are especially crap in those age groups.

Just the other day I was looking at the vax maps in university towns and feeling staggeringly unsurprised (look at Southampton and look up where the halls all are). 

It was the absence of all the leaders and caveats that made it look really bad taken in isolation and at face value. The 'second best...' statement just doesn't hold on its own. But when you phrase it as you did earlier, with staying outside listed in there too I can't disagree with that, and never would have taken issue. Sorry (to everyone) that this has dragged on so long; we could probably have got here much quicker somehow.

Thanks, I'll buy a pint now.

The UK is getting on for 130,000 Covid deaths, Australia has had less than 1,000 with a bit less than 1/2 the population.

The theory that AUS and NZ are going to get it bad too just a bit later is attractive because it makes it possible for Tories/Brexiteers/Lockdown resistors in the UK to avoid facing the fact that they were totally wrong.  

They're not doomed but the keep it out strategy was always incredibly fragile and it was a case of when, not if, it was going to fail in exactly the way it has. It's out there now. TTT won't get control back.

It has bought time to watch and wait, so they'll be starting their pandemic in a world that already has vaccines, although they regrettably haven't used that time to get them in arms.

Dontcha just hate those guys who can’t admit when they’re wrong!

TTT in both countries normally comes alongside a highly restrictive area or state lockdown. Delta has hit Australia hard but NZ hasn't lost control as yet. Luckily The Guardian has an Australian version  and reports on NZ as well

https://www.theguardian.com/world/2021/aug/23/big-questions-new-zealand-cov...

Worth it for First Dog on the Moon on its own.

https://www.theguardian.com/commentisfree/2021/aug/23/thank-goodness-matt-c...

After 128,000 Covid deaths (2/4/21), the UK reached 54 jabs per 100 population.

After 26* Covid deaths, NZ has currently reached 54 jabs per 100 population.

NZ is far ahead (13 months!) of the UK in terms of vaccinating people before they die.

Hopefully this will be of great benefit to them if/when their decisive and so far successful strategy fails.

*equivalent to the UK reaching 54 jabs per 100 population on 23/3/20 when in total 350 people had died of Covid in UK.

Having done so brilliantly in every other aspect, it is still disappointing that the vaccine roll out has been so slow in NZ. To be behind Ecuador, Peru, Paraguay, Morocco, Dominican Republic, Rumania, Albania, Russia, Kazakhstan ... is not hugely impressive. While the immediate urgency for vaccination was clearly less in NZ, the potential cost of delay was always obvious, and is now staring the Kiwis full in the face. I hope they can get their Delta outbreak under control, but there's over a hundred cases of community transmission already, and it's in Wellington now as well as Auckland. I fear the worst. Still, if anyone can, Ardern can.

Many countries are now at around the 130 mark. If NZ had used the time they bought to get in that territory before this happened they'd could have been in a very much more robust position than they are, and would have a much better chance of keeping the death toll down.

One point which explains however bad things become it wont be as bad as the UK (assuming hospitals don't become overwhelmed)  is most of the most vulnerable who want to be jabbed are jabbed. As such their per capita deaths are almost certainly going to end up way lower than the UK. NSW who are ahead of most of Australia, but hit the hardest by delta,  say they can jab everyone in the higher risk groups by early autumn.

How could they have gone much faster? Where does the additional supply of vaccines come from? A lot of the countries listed by Wicamoi are using Russian or Chinese vaccine where crucial health information on effectiveness and side-effect risks is not openly available.

This is a fair point. To what degree has the roll out been constrained by supply, and to what degree has the supply itself been constrained by factors within NZ control? I have no information on that, and I'm not attributing blame - but it's still disappointing that relatively few Kiwis have been vaccinated thus far.

I don't know. Not trying to direct blame. Many of the comparable nations (look at Spain, France, Canada, UK, Germany) have got to the ~130s using the Pfizer and AZ jabs. Some of them (look at Spain) have done the bulk of that relatively recently and quickly. I don't see why NZ is where it is. Whatever happens they'll struggle to kill as many people as we have, but they definitely could have been better placed.

Condescending much! Can I just point out that despite its very restrictive policies and mass vaccination programme, the UK has more deaths per capita from Covid than all but 2 of those countries.

The EU, US and UK have their own manufacturing and have self prioritised. Some other countries have paid a small fortune for small populations, with likely lobbying help on accessing that supply (Israel and Gulf states).

How on earth is "If NZ had used the time they bought to get in that territory before this happened they'd could have been in a very much more robust position than they are," not criticising?

Just checked and all but one of those had Chinese or Russian vaccine alongside other sources. The odd one out is Paraguay where there was a scandal I'd completely missed about potential Taiwan involvement in access to Indian vaccine. I wonder how many of these supply deals have strings attached, especially Chinese vaccine supply, given the history of chinese aid and loans.

On this particular point I agree with LSRH. I think a country with the wealth of Oz or NZ could have done much better on vaccination. They obviously didn't place quite so high a priority on it last year and are now behind as a result. The UK, US and EU have vaccine availability primarily because they put funding in early and committed to big purchases, not because they have manufacturing sites in their territory and have kept it to themselves. The US had lots of AZ that it never used, the UK has had lots of Pfizer, the EU has had a Moderna plant built. AZ has been used in large quantities elsewhere in the world too. Obviously the market size of these places helps them, but I don't believe places like Oz and NZ couldn't have done much better than they have in this regard.

In summer 2020 when the early vaccine agreements were made, these countries still had realistic-seeming expectations of elimination. Alpha hadn't yet come along and many people were still pushing for an elimination strategy in the West. It's wouldn't have been surprising (speculation) if major vaccine purchase deals were quite a long way down NZ's priority list.

I believe  in Aus they have 10 million or so doses of AZ which preople are refusing to have as they want the Pfizer one. This is causing alot of issues and the medics are getting exasperated and saying just get the AZ.

NZ's priority was and is elimination. It looks like the Delta variant is putting paid to that so they are probably going to have to have a rethink.

We will have to agree to disagree. S I understand it, If supply growth had run smoothly and other factors were not involved I'd maybe agree but the initial contracts looked very expensive unless you had a major infection level problem (a big reason we were faster initially than most of the EU). After that everyone was in the queue. By the time delta hit the west it was all too late. The big western states are not hoarding by blocking supply, they have just bought up most of the early supply chain on a delivery schedule and not giving much away (except spare UK AZ). Please feel free to show I'm wrong.

There's an AZ plant in Oz. Oz and NZ agreed on joint AZ manufacture and supply.

Spoke to a friend yesterday, two weeks back from NZ, where she has been for much of the pandemic. She's a few years older than me, but hadn't been vaccinated at all. Got her first (Pfizer) jab soon after arriving here.

NZ have done so many things right, but their vaccine rollout has been poor; they have failed to use the good health grace their earlier policy gave them to roll out a vaccine, and prepare for a return to global unlocking. Friend told me they initially weren't keeping records of who had been vaccinated.

That's true but the Oz government threw a spanner in the AZ works. I apologise as I should have been clearer I'm talking of imported supply of Pfizer, Moderna etc (which is available). As much as I dislike the Australian government they have real problems with hesitancy in their population, not helped by negative press on AZ.

https://www.bbc.co.uk/news/world-australia-56825920

Other countries I could have mentioned - all (just like in the previous list) with substantially lower GDP per capita than NZ, but higher vaccination rates - were Turkey, Argentina, Panama, Saudi Arabia, Mexico, Brazil, Uruguay, Cuba, South Korea, Malaysia, Jordan, Serbia, North Macedonia, Montenegro, Cambodia, El Salvador and Azerbaijan. 

I'll look at most of those but not the Saudis who are above NZ on GP per capita.

Costa Rica seems to be the only other country in South America not using Russian ir Chinese vaccines. Panama are getting major help from the canal company.

https://www.as-coa.org/articles/timeline-tracking-latin-americas-road-vacci...

None yet. S Korea is also a fully developed economy and shouldn't count and the search found another interesting link Id missed before, so thanks.

https://www.wsj.com/articles/israel-south-korea-to-swap-700-000-pfizer-bion...

Another surprise to me, Cuba has developed its own vaccines.

https://www.theguardian.com/world/2021/aug/22/cuba-coronavirus-vaccines-hea...

That's not quite how deliveries were decided, but anyway Canada was in the same position until supply became available to them and look what happened there.

It is criticising. I was criticising their relatively slow vaccine rollout. If they'd used the time they bought to be as vaccinated as eg Canada, then there would be no criticism. They've been incredibly successful in every other respect.

Fair enough - and I'll be interested in your findings. On the wikipedia list I'm looking at Saudi Arabia has less than half NZ's GDP per capita and is ranked 54 to NZ's 28, but let's not quibble.

Weird (but I checked PPP)

https://en.wikipedia.org/wiki/List_of_countries_by_GDP_(PPP)_per_capita

Not quite eh?....this is a reminder of Israel deals.

https://www.theatlantic.com/ideas/archive/2021/01/why-israels-vaccine-succe...

Israel sold out completely, yeah. But that doesn't explain Spain, Canada, all the other Western nations. And domestic manufacture doesn't either.

Malaysia had AZ and Pfiizer supply issues. But also used Sinovac with better supply

https://en.wikipedia.org/wiki/COVID-19_vaccination_in_Malaysia.

So all others using Russian and/or Chinese vaccines.

Plus finally some background and links (too few on Wikipedia) for other country supplies.

https://en.wikipedia.org/wiki/Deployment_of_COVID-19_vaccines

https://covid19.trackvaccines.org/country/

(with a world map with the shittyist readability I've ever seen)

I think it is arguable which of PPP (yours) and nominal (mine) is more relevant in this context, but happy for you to remove Saudi from the list in any case!

I hope it works and they can produce it in bulk without being crippled by the sanctions mentioned in the article.

Taiwan has also developed its own vaccine (edit - I see you knew that).

https://www.reuters.com/world/asia-pacific/taiwan-start-contested-roll-out-...

Hopefully more and more production of safe & effective vaccines will speed things up. Currently a pretty amazing 35M jabs worldwide per day but that's pretty static.

Looking at manufacturing capacity it looks like production is rocketing.

https://en.wikipedia.org/wiki/Deployment_of_COVID-19_vaccines#Manufacturing...

I knew about Taiwan... a highly developed and wealthy nation.

Spain has very low vaccine hesitancy and a well organised public health system.

https://en.wikipedia.org/wiki/COVID-19_vaccination_in_Spain

Canada made deals (and were let down a bit initially) and would be better still without AZ withdrawal.

https://en.wikipedia.org/wiki/COVID-19_vaccination_in_Canada

Which other western nations?

While rich countries can certainly outbid poorer countries to secure vaccine supplies, this is not the only factor on which an effective vaccine programme depends. Russia (on your original list) has way lower GDP than either NZ or the UK but masters nuclear power, space travel, particle accelerators... and vaccine development. This article suggests their vaccine is as effective and safe as western products https://www.nature.com/articles/d41586-021-01813-2 . I'd be interested to learn more about the distribution strategy for both the Russian and Chinese vaccines. They seem to be able to supply a lot of the world's poorer countries.

Another factor is organisational ability. Jordan (on your second list), despite having low GDP, introduced a very effective testing programme and managed to contain both of its infection waves despite having a low rate of vaccination. They also made good use of their initial 6-month isolation policy to build up their healthcare capacity.

NZ's policy of zero-Covid has always had a sell-by-date on it. The delusion and incompetence is quite astounding.

Covid is proving to be yet another illustration of Western complacency and overconfidence. 

Yes. It does.

I'm not sure what you're asking for now. Pick any you want, NZ vaccination is way behind most of them.

Because only Pfizer was approved in NZ until the end of July. You can argue that was a mistake but given that there were real supply problems. Even with AZ there have been supply problems (again see Malaysia's vaccination program).

It's all a bit like you saying people have a choice not to work indoors at work or use public transport (something I see as presenting a completely unreasonable choice to the most disadvantaged ).

Russian and Chinese vaccine distribution strategy is about gaining influence around the world; at least in Russia's case on preference to immunising their own population very quickly. They have also signed deals that they have been unable to deliver on.

Eg https://www.reuters.com/world/americas/argentina-raises-heat-russia-over-sp...

Russian and Chinese vaccine distribution strategy is about gaining influence around the world; at least in Russia's case (and maybe China's?) in preference to immunising their own population very quickly. They have also signed deals that they have been unable to deliver on.

Eg https://www.reuters.com/world/americas/argentina-raises-heat-russia-over-sp...

My strategy these days is just to ignore him. Never tried saying things three times as an alternative

That's a pretty twisted and selective misrepresentation that you've deliberately rephrased to try and make me sound evil. I don't appreciate that. At all.

What I said was that it is not "unavoidable", rather it's avoidable at some cost. I never said the cost was equitable, nor that it was reasonable to expect everyone to make the same choice. I know that is what I said, because I spent a lot of time choosing my words because I knew you would be waiting to pounce and accuse me of something like this if I didn't tread very carefully around it.