For that sat and watched those boring videos I made modelling Covid-19, I reviewed them today and was shocked how accurate the predictions made 7 weeks ago are.
I've done one more video, looking at the predictions and have made further predictions based on Boris' model for lifting lockdown to some extent on Wednesday.
@cameronphillips2000, to hear the audio on your videos I had to watch via my Sky box - and frustratingly I could not find this particular video on it. I am using a Mac and tried in both Safari and Firefox browsers.
Well, the person's real but it's just a made up name, see?
Well, the person's real but it's just a made up name, see?
achilles wrote:
@cameronphillips2000, to hear the audio on your videos I had to watch via my Sky box - and frustratingly I could not find this particular video on it. I am using a Mac and tried in both Safari and Firefox browsers.
Yes, that's odd. Also using Safari on a Mac and no audio if I play the clip embedded in the snowHeads page or if I play it direct in the Youtube website.
Looks like the video uses the Opus codec for its audio, which is not available on the majority of popular browsers and platforms.
Anyway, snowHeads is much more fun if you do.
Anyway, snowHeads is much more fun if you do.
apologies the sound did not upload.
I have redone it and it is uploading now.
The conclusions are as follows:
Whether by fluke or genius - the model predicted that where we are today would be 32000 deaths, about 250 deaths per day and a peak death rate of 1150 deaths per day. This has proved the case.
The model goes on to further predict a total death rate of about 37900 deaths and teh whole thing to be out of the system in the UK in ten weeks from now.
I then change the figures from Wednesday to have a social distance factor at exactly the same as in the first ten days of lockdown when we had semi lockdown (building sites still open , lots of public transport etc,)
This seems to be what will happen from Wednesday, though many maybe won't risk it.
Putting these figures in gives a final death count of 56500, as opposed to 37900. It gives 5 moths before it is out of the system. It gives a maximum deaths per day of 350 though, which does not swamp the NHS.
I will post s link when it has uploaded.
My gut feeling is my first model was ridiculously accurate by luck more than judgement and the further predictions above will be some way off.
I will review in a month and see.
What also intrigues me is that even on an excess deaths basis the German figures are very low https://www.destatis.de/EN/Press/2020/05/PE20_162_12621.html despite the often cited "but they're not counting everything" response when the DE/UK disparity is brought up.
You'll get to see more forums and be part of the best ski club on the net.
You'll get to see more forums and be part of the best ski club on the net.
I think you're probably right. I based my model based on published figures and It predicted inline with published figures. If the actual figures are a certain percent higher then my future predictions will just be at that same ratio
Ski the Net with snowHeads
Ski the Net with snowHeads
Looks interesting - congratulations on both putting it together, and being prepared to put it out there to be reviewed and potentially ridiculed.
The divergence it is likely to suffer is probably going to come from how accurate you can be on the impact of the remaining lockdown on the replication rate. Your changes from 0.3 to 0.6, 0.8 and 1.0 are of course fairly arbitrary, and the reality might be 0.7 or 0.9 instead, but you can always tweak the model so it fits the reality and then read off what the effective rate is that produced that result.
snowHeads are a friendly bunch.
snowHeads are a friendly bunch.
Quote:
The model goes on to further predict a total death rate of about 37900 deaths and teh whole thing to be out of the system in the UK in ten weeks from now.
I would put a fair bit of money (for a good cause....) on the whole thing NOT being out of the system in the UK in ten weeks. Anybody who understands how to "make a book" (which I don't) willing to take this on?
And love to help out and answer questions and of course, read each other's snow reports.
And love to help out and answer questions and of course, read each other's snow reports.
@pam w, 10 weeks?
I'm with you.
I think 10 years from now it may well be in the disease menu and we will just have to learn to live (and sometimes die) with it.
A vaccine would be great, but let's remember the rather similar SARS CV virus has no vaccine and AIDS has no vaccine even after about 20 years of huge effort and money.
So if you're just off somewhere snowy come back and post a snow report of your own and we'll all love you very much
So if you're just off somewhere snowy come back and post a snow report of your own and we'll all love you very much
Hmm - where are SARS, MERS, swine flu et al now? Who is to say that Covid-19 will stay around, rather than flashing into focus and then vanishing like the others? The only way to answer the question is to wait and see. It may stay with us forever, or it may not.
You know it makes sense.
You know it makes sense.
@cameronphillips2000, Can you do one for the lottery?
Otherwise you'll just go on seeing the one name:
Otherwise you'll just go on seeing the one name:
Quote:
It may stay with us forever, or it may not.
Well yes, obviously. But there's a big margin between "ten weeks" and "forever". Flu has stuck around in a big way, and in 2017-18 killed 61,000 in England and Wales (the majority in care homes), led to a million hospitalisations, despite historically high take up of the moderately effective flu vaccinations. What would it be doing without a vaccine?
Poster: A snowHead
Poster: A snowHead
@pam w, seasonal flu is not a single virus, and the vaccine is reckoned to be no more than 60% effective as it is based on a prediction of which strains will be prevalent this time around.
The impact of the vaccine is that people who would otherwise die are kept alive. As we’re now realising more clearly, sometimes that just means they die of something else instead.
Obviously A snowHead isn't a real person
Obviously A snowHead isn't a real person
@cameronphillips2000, thanks for fixing the sound. Interesting video.
Well, the person's real but it's just a made up name, see?
Well, the person's real but it's just a made up name, see?
cameronphillips2000 wrote:
I then change the figures from Wednesday to have a social distance factor at exactly the same as in the first ten days of lockdown when we had semi lockdown (building sites still open , lots of public transport etc,)
This seems to be what will happen from Wednesday, though many maybe won't risk it.
Well done to you for sticking your neck out. I wonder what comes out at certain points between the two 'extents' of lockdown?
We're probably a lot more savvy and accustomed then we were then, plus there's a lot more glove and mask wearing and availability of hand sanitation etc so maybe the difference won't be a stark as that?
As for the 5 months point I did wonder where the chancellor picked the end of October from so maybe your modelling isn't too dissimilar to their current outputs in terms of when it is out of the system?
@cameronphillips2000, very interesting, and really enjoyed your commentary. What were your models showing as the overall infection rate In the population - was the that the ~50 M column?
Also - if you’re taking requests - I’d be really interested to see where your models predicts we’d be now had we shutdown a week or so earlier in line with other EU countries? I’m assuming the numbers would be 5-10 times lower given it was doubling every 2-3 days at the start, but I haven’t seen anyone quoting estimates (until a couple of days ago) - which is surprising given how damning it would be.
Anyway, snowHeads is much more fun if you do.
Anyway, snowHeads is much more fun if you do.
That's a really interesting point. I shall put it into the model and see what happens.
[b]
If you look at the ONS stats, rather than the COVID-specific stats, they suggest that we're already past 50k deaths on an "excess death" basis, which seems realistic as a methodology. On this basis, your model may be undercooking things a bit.
Yes but I think we also need to allow for the possibility that the X number of actual Covid deaths currently not counted as Covid related will be offset by the Y number of non-Covid deaths that are currently counted as Covid related.
We can only guess at the actual values for X & Y but as I understand it there is to be a huge reclassification exercise carried out using a surprisingly large number of new cause of death classifications currently being downloaded from the WHO.
Of course whatever version of the 'truth' we get will probably end up being politically motivated. i.e. if it turns out to be a huge economic disaster relative to the actual loss of life then we should probably expect the Covid death toll to be exaggerated in justification whereas if we dodge the worse cases of financial armageddon then it might go the other way as countries such as the UK strive to manipulate their position in the 'international death league'?
Then you can post your own questions or snow reports...
Then you can post your own questions or snow reports...
What's everyone's take on how a country's previous success in preserving life for the vulnerable feeds into potential death tolls?
IE
A country like the UK has been relatively successful (up until Covid) at preserving life in the very elderly and chronically ill, plus this year in particular we had a very low toll from flu prior to Covid.
whereas
A country like, say Somalia or Niger, has been (I presume) much less successful in preserving life for the vulnerable. So they have less low hanging fruit for the virus to take, plus less (again I presume) diet induced obesity, diabetes etc etc.
whereas
A country like say Japan (in parts certainly) have the preservation of life vis healthcare but also have very long lived individuals due to historic diet/lifestyle factors.
Does that leave them in a position where Somalia won't have the death toll because their population is on average 'stronger' or is that far outwayed by the lack of healthcare infrastructure?
What about a country like Japan with it's high number of the 'super elderly', or conversely all those African countries where nearly half the population is under 15.
In Europe Ireland has 50% more under 15s per head of pop than Italy has, that's got to skew % death tolls (putting aside Italy's unenviable position of first man over the top in Europe).
Apparently your chance of dying if you get Covid doubles every 4-6 years of life, from a very very small risk in the under 5's on up.
So...........wow.......... @cameronphillips2000 model that. Actually I guess not that hard, once you establish a % death toll for any individual country you plug that in, it's the % that's hard to know.
Well done btw, for doing the work, and as others have said for sticking your head above the parapet and putting it out.
After all it is free
After all it is free
robboj wrote:
snowdave wrote:
[b]
If you look at the ONS stats, rather than the COVID-specific stats, they suggest that we're already past 50k deaths on an "excess death" basis, which seems realistic as a methodology. On this basis, your model may be undercooking things a bit.
Yes but I think we also need to allow for the possibility that the X number of actual Covid deaths currently not counted as Covid related will be offset by the Y number of non-Covid deaths that are currently counted as Covid related.
We can only guess at the actual values for X & Y but as I understand it there is to be a huge reclassification exercise carried out using a surprisingly large number of new cause of death classifications currently being downloaded from the WHO.
Of course whatever version of the 'truth' we get will probably end up being politically motivated. i.e. if it turns out to be a huge economic disaster relative to the actual loss of life then we should probably expect the Covid death toll to be exaggerated in justification whereas if we dodge the worse cases of financial armageddon then it might go the other way as countries such as the UK strive to manipulate their position in the 'international death league'?
The relatively "easy" solution (at least statistically) is to look at how many standard deviations from the mean the excess death number is (irrespective of cause). Of course, there could be another disease going round that kills a lot of people, but absent that, this tells us the net effect of COVID, and can't be distorted by politicians or allocations of deaths to COVID or anything else. Hence why when people get all exorcised about whether Germany is reporting its deaths correctly, it doesn't actually matter, because Germany clearly has many fewer excess deaths (for any reason) than we do, so if its not COVID killing people, something else pretty nasty is going around in the UK!
It's possible that there is a bit of a "beneficial" effect in Germany, in that last year (I think) there were a lot of excess deaths from Flu, whereas in the UK we didn't have a bad flu season, so maybe Germany killed a load of its vulnerable people last year instead, but I'm getting into conjecture here.
You'll get to see more forums and be part of the best ski club on the net.
You'll get to see more forums and be part of the best ski club on the net.
snowdave wrote:
The relatively "easy" solution (at least statistically) is to look at how many standard deviations from the mean the excess death number is (irrespective of cause). Of course, there could be another disease going round that kills a lot of people, but absent that, this tells us the net effect of COVID, and can't be distorted by politicians or allocations of deaths to COVID or anything else. Hence why when people get all exorcised about whether Germany is reporting its deaths correctly, it doesn't actually matter, because Germany clearly has many fewer excess deaths (for any reason) than we do, so if its not COVID killing people, something else pretty nasty is going around in the UK!
It's possible that there is a bit of a "beneficial" effect in Germany, in that last year (I think) there were a lot of excess deaths from Flu, whereas in the UK we didn't have a bad flu season, so maybe Germany killed a load of its vulnerable people last year instead, but I'm getting into conjecture here.
Yes I agree but I also wonder to what extent the German health service has shut down its other functions compared to the UK. There is a lot of actual and anecdotal evidence that even in emergencies the NHS is anything but open for business as usual which must be having an effect on top of that of the cancelled treatments of existing serious conditions. I heard of a case last night of lady whom almost died of sepsis that arose from untreated toothache which became an abscess and got worse, she was, in the end, literally, hospitalised just in time.
Ski the Net with snowHeads
Ski the Net with snowHeads
I put figures in to the model to see what would have happened if we'd shut down two weeks earlier - as requested
Do you convert your infection rate to an R0? (or vice versa?)
It's notable that a few London based banks and similar companies (including my own) ran their own analysis of the figures as you did and went to social distancing or remote working 1-2 weeks ahead of the lockdown.
And love to help out and answer questions and of course, read each other's snow reports.
And love to help out and answer questions and of course, read each other's snow reports.
Interesting just how much difference a few days can make.
But if the model is correct the death rate would still be rising.
The max deaths per day in the model (137) is reached on day 223 (mid Sept).
So to achieve that figure we would need to be in tight lockdown for over 6 months.
So if you're just off somewhere snowy come back and post a snow report of your own and we'll all love you very much
So if you're just off somewhere snowy come back and post a snow report of your own and we'll all love you very much
@cameronphillips2000, unless I’ve missed it could you run a what If the lockdown had started a week earlier.
You know it makes sense.
You know it makes sense.
Nice work Cameron. Interesting for those of us who are not from the UK too.
I like the Dolomite vids you have done in the past better though.
FAQ
Poster: A snowHead
- I’d be really interested to see where your models predicts we’d be now had we shutdown a week or so earlier in line with other EU countries? I’m assuming the numbers would be 5-10 times lower given it was doubling every 2-3 days at the start, but I haven’t seen anyone quoting estimates (until a couple of days ago) - which is surprising given how damning it would be.
