Donate to Remove ads

Got a credit card? use our Credit Card & Finance Calculators

Thanks to johnstevens77,Bhoddhisatva,scotia,Anonymous,Cornytiv34, for Donating to support the site

Coronavirus - Modelling Aspects Only

The home for all non-political Coronavirus (Covid-19) discussions on The Lemon Fool
Forum rules
This is the home for all non-political Coronavirus (Covid-19) discussions on The Lemon Fool
scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#482844

Postby scotia » February 26th, 2022, 12:49 am

Another week's data - with parameters and scaling as for last week.
First England

Image

The published deaths and the projected deaths (based on admissions) both continue to decline, but on quite different trajectories. Clearly the simple relation of published deaths being a fraction of past admissions is not working. Strange - since looking back to my first attempt at this simple model in December 2020, and following through with parameter adjustments when new variants and vaccines arrived, there has usually been a good correlation.
Anyway - lets not complain about good news.

And now for Scotland

Image

Given the poorer statistical accuracy, there still seems to be a reasonable correlation between the published and projected deaths - although a slightly reduced multiplier would help. The projected deaths (based on admissions) now seems to be levelling off. Assuming the Scottish population is approximately one tenth of the English population, we see that the ratio of published deaths is is around this value - within the statistical accuracy. And the projected English deaths are now continuing to fall towards the pro rata of Scottish deaths.

So currently good news all round. But what will be the effect of looser restrictions? Hopefully, very little. But I certainly prefer the more cautious approach in Scotland to that in England. This caution may be age related. Both myself and my wife are in the 75+ range - so we should be getting a second booster, probably in April.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#484437

Postby scotia » March 5th, 2022, 1:03 am

Another week's data - with parameters and scaling as for last week.
First England

Image

There is a surprising scatter in the recently published deaths - rather like the effect of public holidays, except there haven't been any. However, encouragingly the the trend still appears to be downwards The projected deaths, based on the admissions remains stubbornly higher - and is now showing a modest upturn. Could this be the effect of the relaxed approach to restrictions?

And now for Scotland.

Image

The projected deaths (based on admissions) is now wobbling around a fairly level trajectory - roughly around the same population pro-rata as in England. However the most recent published deaths appear to have dipped followed by a sharp rise - which takes it well above the English pro-rata value. But with the poor statistical accuracy, its probably better to wait for more data before making any significant pronouncements.

Gersemi
Lemon Slice
Posts: 492
Joined: November 4th, 2016, 3:57 pm
Has thanked: 535 times
Been thanked: 222 times

Re: Coronavirus - Modelling Aspects Only

#484482

Postby Gersemi » March 5th, 2022, 10:26 am

scotia wrote:
There is a surprising scatter in the recently published deaths - rather like the effect of public holidays, except there haven't been any. However, encouragingly the the trend still appears to be downwards The projected deaths, based on the admissions remains stubbornly higher - and is now showing a modest upturn. Could this be the effect of the relaxed approach to restrictions?



Could it be because the admissions figure is simply recording the number of people admitted who test positive for Covid and as the community rate is so high that means that a significant number are testing positive even though they are being admitted for an unrelated reason? The number of people in mechanical ventilation beds continues to fall, they are now at the level last in June 2021, before the Delta wave really hit. Cases at that time were well under 10,000 per day. The increase in the number of infections is surely due to the final lifting of legal restrictions - I'm only surprised that people are still getting tested in such large numbers.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#484520

Postby scotia » March 5th, 2022, 1:07 pm

After I submitted the latest graphs, I had a look at the previous week's graphs, and for Scotland I noted an obvious anomaly. Looking back to the previous week, there was a difference in the admission points (projected deaths) plotted at the the same time as in this week's graphs. The admissions (at the same date) seem to have grown! So I looked at the data which I downloaded from the government site, both this week and last week - and there was a significant difference at the same dates! What's going on? Then I found that re-admissions are now being counted, and they seem to have back-corrected their data - how far back I don't know. They suggested that the Scottish data is now in synchronism with the English data - which was apparently corrected at some previous time - but when, and whether it was back-corrected, I don't yet know. The corrections to the Scottish Data are significant. E.G. the admissions for the week ending 21/2/22 have "grown" from 671 to 811 - about 20%.
I think I need to take a good look at the data - fortunately I have kept a list of extracted data used for the past plots, so I should be able to work out what has happened, and when.
Maybe I'm going to sound a bit fussy - but I really feel that the original data should have been left "unadjusted", and a new data set with the backdated adjustments should have been separately published.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#487568

Postby scotia » March 19th, 2022, 1:53 am

Most recent data- with parameters and scaling as for two weeks ago
First England
Image

There is little doubt that the admissions are now climbing significantly, and given the reports on increasing infections, it look likely that the rise will continue. However, encouragingly the deaths are not (yet) showing any significant rise, and the ratio of deaths to projected deaths (i.e. delayed admissions) is at a historically low level. Will this continue?

Now to Scotland

Image

The admissions (projected deaths) are on the way up - but are around a similar level (per person) to those in England. However the published deaths are much closer to the projected deaths than in England. I understand that most of the deaths are in the over eighty cohort, and the number is similar to that which would be expected from Flu. However since I am in my late seventies, I don't get too much comfort from that info.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#489253

Postby scotia » March 25th, 2022, 10:42 pm

With the rapid increase of the latest BA.2 variant, I thought it might be useful to look again, after only a week.
So here's the most recent data- with parameters and scaling as for last week

First England
Image

This looks mildly encouraging. The projected deaths (i.e. time delayed admissions) are now rising steeply, however the published deaths are now well below the projections, and are only increasing slowly. If this pattern of a low death rate were to continue, then I could probably understand why the English government is relatively unconcerned, and has apparently abandoned most restrictions.

However, lets look at the Scottish Data - since the outbreak of the new Omicron BA.2 variant has hit Scotland earlier and harder than in England.

Image

Even allowing for the much poorer statistical accuracy, the Scottish curves are quite different from the English curves. The Projected deaths (delayed admissions) have increased significantly, but now may (optimistically) have peaked. The worrying feature is that the published deaths have suddenly climbed to meet it.
So given that I have plotted the Scottish and English data with a scaling that approximately balances the 1 to 10 population numbers, its immediately obvious that although the admissions are similar in size, the published deaths are significantly different, both in actual size and in the rapidity of the increase. Whereas the English data was encouraging - could it be a false dawn? Is it possible that the latest Omicron BA.2 variant has a sting in the tail - with a sharp increase in delayed deaths which is yet to appear in the English data? I hope not. But the current evidence would make me more cautious about removing restrictions.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#490971

Postby scotia » April 1st, 2022, 11:49 pm

A further week of data, with the same parameters, but with the scaling adjusted to keep the displays within bounds. The 10:1 ratio of scaling between Scotland and England has been maintained.
First England

Image

The projected deaths (i.e. based on admissions) continue to rise, with the published deaths also rising - but with a significantly lower gradient than the projected deaths (admissions). The factor of 0.15 is now clearly too large - possibly by a factor of 2. So it looks like the current outbreak is resulting in a much smaller death rate. The vaccinations will certainly be responsible for part of this, but also it may now be due to the younger sections of the population who are getting back to normal activities, and are more likely to be infected with this latest variant, whereas the oldies are still trying to keep virus free - waiting for the next booster.

And now Scotland

Image

Even allowing for the much poorer statistical accuracy, it looks like the projected deaths (admissions) and published deaths have now peaked. However the correlation is quite unlike that in England - the published deaths are much closer to the projected deaths than in England - the factor of 0.15 seems reasonable. The projected deaths in England and Scotland (based on admissions) are similar on a population (10:1) basis. However the published deaths are significantly different. The Scottish published deaths per week has peaked around 200. The English published deaths over the past week is 820 - and rising. If it were to follow the Scottish pattern , it would peak at around 2000. It currently does not appear that this is likely.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#493174

Postby scotia » April 9th, 2022, 8:14 pm

A further week of data, with the same parameters and scaling as last week
But first I'll look at Scotland

Image

We seem to be over the peak in projected deaths (i.e. admissions), and within the poor statistical accuracy, the published deaths seem to be following similarly. There was a steep climb in the published deaths up to the peak, a feature that was not previously seen in England, where the published deaths remained well below the projected deaths, and showed no signs of shooting upwards. So lets look at England now

Image

It would appear that the projected deaths (i.e. admissions) have probably reached a peak - but its the last 5 days of published deaths that are ringing my alarm bells. These weekly totals have climbed from 820 to 1295, and the gradient appears to be increasing. Last week I predicted
The English published deaths over the past week is 820 - and rising. If it were to follow the Scottish pattern , it would peak at around 2000. It currently does not appear that this is likely.

Its early days, but I am no longer so convinced.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#494600

Postby scotia » April 16th, 2022, 1:03 am

A further week of data, with the same parameters and scaling as last week - except there is only 6 days of new data ending on 14th April (in both Scotland and England), with no data published on Easter Friday 15th April. I don't know if there will be data published on Easter Monday - so we may see data skewed over the coming week.

First I'll look at Scotland
Image

within the statistical accuracy, the projected deaths (based on delayed admissions) and the published deaths seem both to have reached their peaks, and are now declining. A slightly lower multiplier than the 0.15 would give a better match to the two graphs.

And now for England
Image

The projected deaths (based on delayed admissions) now seem to have peaked - and are falling gradually. The published deaths which fell well below the projected deaths are now climbing rapidly (and erratically) towards the projected deaths. Could there be a social or a medical reason (or both) for this steep rise? A possible medical reason could be that the most recent variant of Omicron is proving more fatal - and a social reason may be that the elderly and most vulnerable are finding it more difficult to avoid infection due to the loosening of protective rules.

Julian
Lemon Quarter
Posts: 1385
Joined: November 4th, 2016, 9:58 am
Has thanked: 532 times
Been thanked: 676 times

Re: Coronavirus - Modelling Aspects Only

#494767

Postby Julian » April 17th, 2022, 11:23 am

scotia wrote:A further week of data, with the same parameters and scaling as last week - except there is only 6 days of new data ending on 14th April (in both Scotland and England), with no data published on Easter Friday 15th April. I don't know if there will be data published on Easter Monday - so we may see data skewed over the coming week.

One thing occurs to me that might be making the modelling more challenging now. It’s been pretty widely stated for quite a few weeks that a significant percentage of people in hospital with Covid-19 were admitted for something else rather than Covid-19 being the primary condition. I haven’t seen any actual data on that although one would hope it is available (admittedly I haven’t really looked) but I’ve heard figures in the 40% to 50% range mentioned by some commentators e.g. John Campbell.

Let me say straight up that this is not necessarily a suggestion for you since the necessary reliable extra data is probably very hard to come by or in some cases might not even exist, and you already put in a lot of effort on this anyway, but in theory what the above suggests to me is that a refinement on the model might be to split the hospital admissions into two groups, those admitted specifically for C-19 (the “C19s”) and those admitted for other reasons that happened to incidentally test positive for C-19 (the “Others”) so that different weightings could be applied to each group.

Assuming that the death figures you are using are the within-28-day numbers then it is very likely that many of the deaths in the Others group will still feed into the death numbers (but not all since I believe the within-28-days is still talking about first positive test and the test on admission might not be the first positive) but I would not be at all surprised if the weightings for the two cohorts behaved differently over time. For instance I would expect to see the weighting for the C-19s to have been decreasing over time as treatment/therapeutics have improved whereas at certain stages in the pandemic including recently I wouldn’t be surprised to see the weighting for the Others to have been increasing over certain periods of NHS overload due to people presenting later, ambulance shortages causing longer delays before stroke, heart attack and trauma victims can be got to A&E and probably other factors too.

BTW, when I say “weighing” above I mean the scaling factor you are using to predict what percentage of any given days admissions will end up dying.

- Julian

Itsallaguess
Lemon Half
Posts: 9129
Joined: November 4th, 2016, 1:16 pm
Has thanked: 4140 times
Been thanked: 10023 times

Re: Coronavirus - Modelling Aspects Only

#494774

Postby Itsallaguess » April 17th, 2022, 12:06 pm

Julian wrote:
scotia wrote:
A further week of data, with the same parameters and scaling as last week - except there is only 6 days of new data ending on 14th April (in both Scotland and England), with no data published on Easter Friday 15th April. I don't know if there will be data published on Easter Monday - so we may see data skewed over the coming week.


One thing occurs to me that might be making the modelling more challenging now.

It’s been pretty widely stated for quite a few weeks that a significant percentage of people in hospital with Covid-19 were admitted for something else rather than Covid-19 being the primary condition.

I haven’t seen any actual data on that although one would hope it is available (admittedly I haven’t really looked) but I’ve heard figures in the 40% to 50% range mentioned by some commentators e.g. John Campbell.


Might it get to be even more nuanced than that, as we get to fewer community transmissions, due to the relatively high number of people still actually catching the virus in hospital after being admitted for a different 'primary condition'?

What I'm getting at is that if broader transmission rates are more difficult to bring down in a hospital setting, when compared to a natural waning of community transmission outside of that healthcare setting, then are we likely to see the ratio of deaths-to-cases rise because of that 'caught-in-hospital' susceptibility as well?

One of the great mysteries of the past couple of years is how we've not had better visibility of the number of in-hospital transmissions, never mind got to grips with keeping them lower...

Cheers,

Itsallaguess

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#494783

Postby scotia » April 17th, 2022, 12:51 pm

Thanks for the comments - Julian and Itsallaguess
I have been getting a bit concerned about the quality of the (English) data on deaths by published date. Its almost like there is now a low priority about providing that data - perhaps a feeling that Covid is past, and we should be getting on with other matters. Could the teams collecting such data have been (partially) re-located?
Even the 7 day average values of these deaths by published date are now jumping around with amounts well in excess of many standard deviations. Why? Its particularly annoying, since we seem to be looking at a very pronounced rise in such deaths - at a time when other statistics seem more favourable. And I fear that the Easter holiday days are going to further muddy the waters.
I have just returned from a holiday, so my garden needs significant attention. But I'll try to make time to have a look at other available statistics.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#496098

Postby scotia » April 24th, 2022, 12:28 am

I downloaded Friday's (22nd April) data, and plotted my usual graphs. I had expected some anomalies due to the 4-day Easter weekend, although the seven day averaging should have ironed out the major fluctuations - but I was disappointed. After some thought I have decided to display the graphs - but it looks like the published deaths provides little information on the direction of their progress. Scaling and parameters as previous week.
First England

Image

I get the feeling that publishing the deaths is now a low priority, and they may be publishing them only when it is convenient. Hence the uncorrelated jumps in the weekly numbers. So I'm afraid its difficult to extract any pattern from them. The admissions do not suffer from this problem, and they have definitely peaked. Assuming that the actual deaths follow the delayed admissions - then it appears that the projected deaths are currently near a peak, and may be declining.

And now for Scotland

Image

The statistics are poor - but there appears to be a recent jump in the published deaths - which is probably due to the Easter weekend effect. So possibly both the projected deaths (based on admissions) and published deaths are now falling. A weakness in the Scottish data is the delay in publishing admissions. In the 22/4/22 data, the last admissions data was at 15/4/22.

dealtn
Lemon Half
Posts: 6072
Joined: November 21st, 2016, 4:26 pm
Has thanked: 441 times
Been thanked: 2324 times

Re: Coronavirus - Modelling Aspects Only

#496221

Postby dealtn » April 24th, 2022, 10:26 pm

Am I missing something?

You have published a series of numbers over around 50 days. The red being the predicted deaths of the model, and the blue being the actual deaths. For the entirety of the period the red is higher than the blue. Are we not to simply conclude the model is wrong and overly pessimistic? I hope that's not too controversial but I can't see what is being attempted here. Apologies if I have got this wrong, or you are trying to show something else, as you have clearly gone to a lot of effort.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#496331

Postby scotia » April 25th, 2022, 12:46 pm

dealtn wrote:Am I missing something?

You have published a series of numbers over around 50 days. The red being the predicted deaths of the model, and the blue being the actual deaths. For the entirety of the period the red is higher than the blue. Are we not to simply conclude the model is wrong and overly pessimistic? I hope that's not too controversial but I can't see what is being attempted here. Apologies if I have got this wrong, or you are trying to show something else, as you have clearly gone to a lot of effort.

These publications have been going on over long periods, roughly at 1-2 week intervals. Initially, on looking at the English data (since the Scottish data is statistically poor), it was interesting to find a reasonably close multiplicative relation between the published 7 day Covid deaths and the 7 day Covid admissions of a number of days previously over a substantial period - and the two parameters (multiplier and days previously) were optimised for best fit on that initial data. Over a considerable period (pre-vaccinations) this initial fit was surprisingly accurate, even during various lockdown periods. It did slightly diverge, and the parameters were amended to provide a better fit - and in hindsight this was probably the effect of the new delta variant. Then came the vaccinations, and the change was dramatic, with a substantial fall in the admissions, and even larger falls in the deaths. This eventually settled down to approximately fit the current parameters, which appeared to work until the beginning of this year (when Omicron became significant). Looking at the submission of January 14th, we see a sharp rise in the admissions, and yes, it appears that the multiplicative factor is becoming over-pessimistic, although the quality of the deaths statistics has become poor (jumps of multiple standard deviations). Now move on to the submission of February 4th - and we see that although the published deaths peak fell well below the predicted deaths - it continued for a longer time, until they came back into reasonable agreement. Then jumping forward to the submission of March 5th, we see that after a short period of agreement, the published deaths fell much faster than the projected deaths. And then looking at the April 1st submission, we see that the projected deaths (based on admissions) are now rising, but the published deaths are only barely rising. At this stage I might have proposed that this is definitely an OmIcron effect - it is much less life-threatening, and I should possibly substantially reduce my multiplicative factor. But I decided to hold fire, since there was some more disturbing information from the Scottish data where Omicron had struck sooner.
So back to the last submission. There is evidence that there has been a sharp rise in the published deaths - but the poor quality of this data does not allow us to speculate whether or not this has peaked, and what has caused it. Is it social or medical or administrative? I think I'll wait for more data before making predictions. But yes - I'm hopeful that the current multiplicative factor is too pessimistic.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#497716

Postby scotia » April 30th, 2022, 5:16 pm

Another week's data (to Friday 29th April)
First England

Image

And once again we get jumps in the published deaths up and down over many standard deviations. OK - the 4-day Easter Holiday will have an effect - but I'm beginning to think that there is now little attention being paid to publish the deaths at the most accurate time. So with data like this - what observations can be made? I think it is unlikely that the double peak is real - and probably we should think of it being a single peak, probably further time shifted by the publications catching up with the holiday data. And probably the multiplier should be lower. But I wouldn't like to submit such conclusions to scientific scrutiny. And next week there is a Bank Holiday which will further mess up the published deaths.

Now on to Scotland - with much poorer statistics

Image

Again - I'll assume the apparent double published deaths peak is caused by the reasons I have suggested for England.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#499175

Postby scotia » May 7th, 2022, 11:16 pm

Another week's data (to Friday 6th May)
First England
Image
The published deaths continue their erratic behaviour, which I'll generously suggest is due to public holidays - but which I suspect is now an increasingly low priority in the data collection. But if we are inclined, in a rather optimistic manner, we could assume a single peak which is now declining at a slightly lower level than the projected deaths. But it doesn't explain how before this peak, it was substantially below the projected deaths, then made a sudden increase upwards. As I have pondered Is this behaviour a social or medical or administrative phenomenon? I.E could it have been oldies venturing out believing the epidemic was over (social), or was it a new, and more fatal version of Omicron (medical), or was it simply erratic collection of the deaths by publish date (administrative). Maybe we will see some stability in the coming weeks to provide a more constant model.

And now Scotland
Image
Allowing for the poor numerical statistics, the published and projected deaths now seem to be falling at a similar rate.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#500521

Postby scotia » May 14th, 2022, 11:03 pm

Another weeks data as extracted late on 14/5/22. For England this contains data to 13/5/22, for Scotland it contains data to 12/5/22
First England

Image

I'm still waiting for some stability in the deaths by publish date - and my comments of last week are still valid.
However i have been playing around with the parameters, and in particular I have been increasing the time difference between the published deaths and projected deaths from 11 to 18 days in order to get the (ill defined) published deaths peak to coincide with the projected death peak. But why? Is it simply due to a deterioration in the provision of the published death data? I think I'll wait to see if some order returns to the published deaths data before attempting a more reasonable fit to the model parameters.
However the good news is that the published and projected deaths both now appear to be falling.

And now for Scotland

Image

Allowing for the poor numerical statistics, the published and projected deaths still seem to be falling at a similar rate

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#502307

Postby scotia » May 23rd, 2022, 1:24 pm

Another weeks data as extracted earlier today. For England this contains data to 20/5/22, for Scotland it contains data to 19/5/22
Frst England

Image

Both the projected deaths (based on admissions) and the deaths by publish date now seem to be continuing on downward curves. The large swings in deaths by publish date are now less evident. I'm now of the opinion that they were administrative - possibly less care was being taken as to when these were entered into the published statistics. On very little data, the recent sharper fall in the deaths by publish date may support the view that there there is now a longer time slip between admissions and these deaths - but whether this is a real medical effect, or is simply due to increased delays in the publications, I don't know. I'll wait for more data.

and now for Scotland

Image

There is a continuing fall in both published and projected deaths, which are in agreement within the poor statistical accuracy.

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#503772

Postby scotia » May 30th, 2022, 2:01 pm

Another weeks data as extracted earlier today. For England this contains data to 27/5/22, for Scotland it contains data to 26/5/22
First England

Image

Both the projected deaths (based on admissions) and the deaths by publish date are continuing on downward curves. The large swings in deaths by publish date are now less evident. I remain of the opinion that they were administrative - possibly less care was being taken as to when these were entered into the published statistics. On very little data, the recent sharper fall in the deaths by publish date may support the view that there there is now a longer time slip between admissions and these deaths. As I stated last week, this could be a real medical effect, but it could also be due to increased delays in the death publications. However I tried out some changes to the two model parameters - I increased the time slip to 18 (from 11) days , and reduced the multiplicative factor from 0.15 to 0.1. I also increased the plotting range from 60 to 80 days - in order to see the peak in the predicted deaths (based on admissions).

Image

I don't think I would draw too many (medical) conclusions from the graph - I suspect I am just trying to model the administrative delays.

And now for Scotland (with last week's parameters)

Image

There remains to be a continuing fall in both published and projected deaths, which are in agreement within the poor statistical accuracy.

Getting back to the quality of the deaths by published date - it is also apparent in the Scottish data, since over the past fortnight there have only been non-null entries in two days out of seven. I can understand the weekend nulls - but not the other days.

I'm happy that we are now seeing falls in admissions and in deaths - but I don't think that this simple 2-parameter deaths model is now producing much predictive information. I'll continue to watch the data, and if anything new of interest crops up, I'll publish the graphs - but I'll drop the current weekly displays.


Return to “Coronavirus Discussions”

Who is online

Users browsing this forum: No registered users and 5 guests