COVID-19

Discussions about serious topics, for serious people
Locked
KAJ
Fuzzable
Posts: 310
Joined: Thu Nov 14, 2019 5:05 pm
Location: UK

Re: COVID-19

Post by KAJ » Sun Nov 15, 2020 4:35 pm

Oh :| . I may have to hand in my cock-eyed optimist card. On the latest UK figures I think I am maybe perhaps beginning to see, after two or three weeks flat, suggestions of a rise in the rate of cases (by specimen date) and of hospital admissions. Ten days into the lock-down I was hoping to see declines.
Screenshot 2020-11-15 at 16.28.05.png
Screenshot 2020-11-15 at 16.28.05.png (15.72 KiB) Viewed 2925 times
Or perhaps it's just the rain making me pessimistic.

User avatar
Bird on a Fire
Princess POW
Posts: 10137
Joined: Fri Oct 11, 2019 5:05 pm
Location: Portugal

Re: COVID-19

Post by Bird on a Fire » Sun Nov 15, 2020 4:43 pm

JQH wrote:
Sun Nov 15, 2020 3:55 pm
Bird on a Fire wrote:
Sun Nov 15, 2020 2:35 pm
Ah yes, leftism is when the government does things, and the more things the government does the more left it is.
Certainly seems to be the Trumpist attitude. I can't see how else they get to believing Joe Biden is a socialist.


Which makes me wonder what they would have called Bernie Sanders if he'd got the nomination.
They're certainly running out of rhetorical runway.

Mind you, even the Democrats didn't seem to realise that Sanders was just proposing normal European policies rather than anything radically leftist in a developed-world context.

I'm not normally a fan of intrusive government (despite being leftish) but dealing with pandemics does seem to be one area where top-down imposition of rules that, regrettably, infringe people's civil liberties is actually quite useful for saving lives. You don't have to go full-on totalitarian about it, but Oz and NZ seem to have done a decent job compared to other wealthy island nations I could mention ;) (How's that for a re-rail?)
We have the right to a clean, healthy, sustainable environment.

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 5:03 pm

KAJ wrote:
Sun Nov 15, 2020 4:35 pm
Oh :| . I may have to hand in my cock-eyed optimist card. On the latest UK figures I think I am maybe perhaps beginning to see, after two or three weeks flat, suggestions of a rise in the rate of cases (by specimen date) and of hospital admissions. Ten days into the lock-down I was hoping to see declines.
Screenshot 2020-11-15 at 16.28.05.png
Or perhaps it's just the rain making me pessimistic.
I actually think that we're near or maybe over the peak as far as cases are concerned


Image
And that the rise in deaths might be starting to slow too. Still rising though.
Image

Again this is the ECDC by date reported, with centred 7-day averages in black so the date is the morning after the UK date (e.g. the 25th September was reported around 4pm on 24th on the UK COVID dashboard)
https://www.ecdc.europa.eu/en/publicati ... -worldwide
Have you considered stupidity as an explanation

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 5:32 pm

I'm getting tired of people denying that lockdowns reduce the infections.

I still think that Israel is the poster child for getting the first wave under control with increasing restrictions, relaxing too soon (at about 10 cases per day) and then needing a second lockdown to cause cases and deaths to fall again.

Image


But this is "That is associational twaddle - period."


https://twitter.com/FatEmperor/status/1 ... 9621677060

It's some weird form of misplaced ideal of proof that their opponents need but where assertions are just fine for them
Have you considered stupidity as an explanation

User avatar
shpalman
Princess POW
Posts: 8241
Joined: Mon Nov 11, 2019 12:53 pm
Location: One step beyond
Contact:

Re: COVID-19

Post by shpalman » Sun Nov 15, 2020 5:45 pm

I mean, the correct answer to "lockdown isn't working because cases* aren't going down" is to lockdown more, or at least to actually try following the lockdown measures. Or at least at least to note that numbers of new cases per day would still be going up exponentially, and they aren't.

* - number of new cases per day
having that swing is a necessary but not sufficient condition for it meaning a thing
@shpalman@mastodon.me.uk

User avatar
sTeamTraen
After Pie
Posts: 2550
Joined: Mon Nov 11, 2019 4:24 pm
Location: Palma de Mallorca, Spain

Re: COVID-19

Post by sTeamTraen » Sun Nov 15, 2020 5:49 pm

Sue Cook, former presenter of Crimewatch, has gone full frootloop.
Something something hammer something something nail

Hunting Dog
Fuzzable
Posts: 264
Joined: Tue Nov 12, 2019 7:48 pm

Re: COVID-19

Post by Hunting Dog » Sun Nov 15, 2020 7:03 pm

sTeamTraen wrote:
Sun Nov 15, 2020 5:49 pm
Sue Cook, former presenter of Crimewatch, has gone full frootloop.
I really can't understand how she's posted that... the pic:
Image
shows deaths going up so wtf is she on about with the test results going up and not the deaths?
Attachments
deaths.jpg
deaths.jpg (75.93 KiB) Viewed 2857 times

User avatar
bob sterman
Dorkwood
Posts: 1123
Joined: Mon Nov 11, 2019 10:25 pm
Location: Location Location

Re: COVID-19

Post by bob sterman » Sun Nov 15, 2020 7:05 pm

sTeamTraen wrote:
Sun Nov 15, 2020 5:49 pm
Sue Cook, former presenter of Crimewatch, has gone full frootloop.
She appears to have fallen under the spell of Sikora, Heneghan and that diet guru bloke off Youtube...

Ex-BBC star Sue Cook, 71, accuses broadcaster of unbalanced Covid coverage and pushing too hard to shut down UK by forever 'wheeling out' 'Professor Lockdown' Neil Ferguson over other scientists
https://www.dailymail.co.uk/news/articl ... erage.html

The diet guru bloke (Cummins) is described as "Chief Program Officer for Irish Heart Disease Awareness" - which sounds very impressive. Except Irish Heart Disease Awareness has the same address as the one Cummins sells his diet guru books from.

User avatar
jdc
Hilda Ogden
Posts: 1925
Joined: Wed Sep 25, 2019 4:31 pm
Location: Your Mum

Re: COVID-19

Post by jdc » Sun Nov 15, 2020 7:05 pm

This was one of the things she was approvingly tweeting about: https://thehighwire.com/videos/pfizer-c ... ne-frenzy/

It's really quite something.

KAJ
Fuzzable
Posts: 310
Joined: Thu Nov 14, 2019 5:05 pm
Location: UK

Re: COVID-19

Post by KAJ » Sun Nov 15, 2020 7:59 pm

jimbob wrote:
Sun Nov 15, 2020 5:03 pm
KAJ wrote:
Sun Nov 15, 2020 4:35 pm
Oh :| . I may have to hand in my cock-eyed optimist card. On the latest UK figures I think I am maybe perhaps beginning to see, after two or three weeks flat, suggestions of a rise in the rate of cases (by specimen date) and of hospital admissions. Ten days into the lock-down I was hoping to see declines.
Screenshot 2020-11-15 at 16.28.05.png
Or perhaps it's just the rain making me pessimistic.
I actually think that we're near or maybe over the peak as far as cases are concerned


<snip chart>
And that the rise in deaths might be starting to slow too. Still rising though.
<snip chart>

Again this is the ECDC by date reported, with centred 7-day averages in black so the date is the morning after the UK date (e.g. the 25th September was reported around 4pm on 24th on the UK COVID dashboard)
https://www.ecdc.europa.eu/en/publicati ... -worldwide
Yes, concentrating on more recent data makes more recent changes clearer. I think my post of Thursday 7:08 pm (based on data.gov.uk) strongly suggested that cases had plateaued. This chart of todays data on deaths by date of death ...
Deaths.png
Deaths.png (5.96 KiB) Viewed 2836 times
... strongly suggests deaths also have plateaued/are plateauing.

I'd expected/hoped the current lock-down to cause a decline in cases about now. In my post of 4:35 today a quick eyeball made me think numbers might in fact be rising - perhaps due to pre-lockdown splurges. A closer look suggests my pessimism is unfounded.

Here I've fitted a quadratic (to reflect the plateauing) with a day-of-week factor.
XYplot2.png
XYplot2.png (16.42 KiB) Viewed 2836 times
The fit is as good as could be reasonably expected, the plateau is clear and well fitted, the quadratic and day-of-week terms are well justified.

Code: Select all

Residuals:
     Min       1Q   Median       3Q      Max 
-0.18239 -0.05675 -0.02204  0.06101  0.18769 

Coefficients:
                        Estimate Std. Error t value Pr(>|t|)    
(Intercept)              9.75039    0.03614 269.764  < 2e-16 ***
poly(date, 2)1           2.11798    0.09598  22.066  < 2e-16 ***
poly(date, 2)2          -0.90577    0.09606  -9.429 8.01e-12 ***
weekdays(date)Monday     0.20610    0.05108   4.035 0.000233 ***
weekdays(date)Saturday  -0.23465    0.05101  -4.600 4.02e-05 ***
weekdays(date)Sunday    -0.26081    0.05104  -5.110 7.87e-06 ***
weekdays(date)Thursday   0.06649    0.04960   1.340 0.187464    
weekdays(date)Tuesday    0.12698    0.05114   2.483 0.017221 *  
weekdays(date)Wednesday  0.13979    0.05122   2.729 0.009307 ** 
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.09542 on 41 degrees of freedom
Multiple R-squared:  0.9496,	Adjusted R-squared:  0.9397 
F-statistic: 96.52 on 8 and 41 DF,  p-value: < 2.2e-16
More reassuringly, the residuals don't show any sign of structure, suggesting my fear of a rise after the plateau isn't justified by the data :D - let's hope to see a downward trend in the next few days.
resids.png
resids.png (10.03 KiB) Viewed 2836 times

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 9:03 pm

KAJ wrote:
Sun Nov 15, 2020 7:59 pm
jimbob wrote:
Sun Nov 15, 2020 5:03 pm
KAJ wrote:
Sun Nov 15, 2020 4:35 pm
Oh :| . I may have to hand in my cock-eyed optimist card. On the latest UK figures I think I am maybe perhaps beginning to see, after two or three weeks flat, suggestions of a rise in the rate of cases (by specimen date) and of hospital admissions. Ten days into the lock-down I was hoping to see declines.
Screenshot 2020-11-15 at 16.28.05.png
Or perhaps it's just the rain making me pessimistic.
I actually think that we're near or maybe over the peak as far as cases are concerned


<snip chart>
And that the rise in deaths might be starting to slow too. Still rising though.
<snip chart>

Again this is the ECDC by date reported, with centred 7-day averages in black so the date is the morning after the UK date (e.g. the 25th September was reported around 4pm on 24th on the UK COVID dashboard)
https://www.ecdc.europa.eu/en/publicati ... -worldwide
Yes, concentrating on more recent data makes more recent changes clearer. I think my post of Thursday 7:08 pm (based on data.gov.uk) strongly suggested that cases had plateaued. This chart of todays data on deaths by date of death ...Deaths.png
... strongly suggests deaths also have plateaued/are plateauing.
Unfortunately there's a huge lag in that data:

Image
Have you considered stupidity as an explanation

badger
Fuzzable
Posts: 264
Joined: Tue Nov 12, 2019 3:23 pm

Re: COVID-19

Post by badger » Sun Nov 15, 2020 9:28 pm

sTeamTraen wrote:
Sun Nov 15, 2020 5:49 pm
Sue Cook, former presenter of Crimewatch, has gone full frootloop.
Just went on a Twitter surf off that tweet. Jesus f.cking Christ. There's no flu because it's all Covid. Half the Covid is false positive. It's just a "bad year" of flu. It's all a conspiracy (yeah, because Govts just love a lockdown, right?). I know soc media amplifies the crazy, but it's f.cking depressing and worrying when the press pick up on it uncritically.

There's definitely a conversation to be had about surviving lockdowns, about care of the elderly, about psychological and other impacts, and the Govt messaging and engagement has been and continues to be pisspoor, which fuels some of this stuff, but f.ck me the unwillingness to see some very simple cause and effect is lunacy.

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 9:51 pm

Well said, Badger
Have you considered stupidity as an explanation

Chris Preston
Snowbonk
Posts: 529
Joined: Tue Nov 12, 2019 8:05 am

Re: COVID-19

Post by Chris Preston » Sun Nov 15, 2020 9:59 pm

After months of no cases of community transmission in South Australia, there are 17 cases confirmed in the past 2 days, 15 of them in one extended family. Original source was a returning overseas traveler who infected a worker at one of the quarantine hotels.

This demonstrates just how quickly numbers can grow from a single infected person. Even when an effective track and trace system is in place.

No doubt some complacency has set in in South Australia due to the long period of no community transmission and the relative freedom that has brought. If any new cases arise today, we can expect a tightening of restrictions.
Here grows much rhubarb.

User avatar
lpm
Junior Mod
Posts: 5944
Joined: Mon Nov 11, 2019 1:05 pm

Re: COVID-19

Post by lpm » Sun Nov 15, 2020 10:12 pm

My suspicion is that the UK has:

- hotspot areas that were in tier 3, and to a lesser extent tier 2, where cases have fallen very successfully - more compliance, in place longer

- non hotspot areas that were tier 1 where cases are rising - poor compliance, not been locked down for long

My own area had grown from very low to what would be tier 2 over the last month.

It doesn't look like 2 more weeks of lockdown is anything like enough to get Covid knocked back. R=1 doesn't get us anywhere and there's no sign R is at 0.7 nationally.
⭐ Awarded gold star 4 November 2021

KAJ
Fuzzable
Posts: 310
Joined: Thu Nov 14, 2019 5:05 pm
Location: UK

Re: COVID-19

Post by KAJ » Sun Nov 15, 2020 10:26 pm

jimbob wrote:
Sun Nov 15, 2020 9:03 pm
Unfortunately there's a huge lag in that data:

Image
I don't think there is in the way you suggest. The data reported today comes from specimens taken one or more days earlier. For example, here's the head of today's data set as I've downloaded it:
Screenshot 2020-11-15 at 21.41.07.png
Screenshot 2020-11-15 at 21.41.07.png (28.3 KiB) Viewed 2764 times
There were 24,962 cases reported today (PubCases) but none are shown for today under SpecCases (by specimen date). That doesn't mean those 24,962 cases are missing from SpecCases, they're just shown with the specimen date, not the date of publication. The missing values under SpecCases for the last few days will be filled in when those results become available, at the same time as they are reported. For a given date the numbers in PubCases will not change, the numbers in SpecCases will - but this reflects delays common to both columns but "hidden" in PubCases.

There is a not-very-predictable delay between specimen date and reporting date, but that delay applies equally to both columns of data. Using PubCases ignores that delay, using SpecCases corrects for it.

The latest few days in SpecCases tend to have silly low values, many results from specimens taken on those days are not yet available. When results are available they appear in (different rows of) the SpecCases and PubCases columns at the same time. Because recent SpecCases are probably incomplete, I tend to exclude the latest 3 or 4 days from my analyses.

Associating any datum with the reporting date builds in and ignores the not-very-predictable delay between specimen date and reporting date, this is time-axis noise whose effect is very difficult to reduce. Associating the datum with specimen date means the delay is explicit in the missing data for recent days. As shown in my recent post this allows one to fit and study time effects with very simple models and little unfitted noise.

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 10:36 pm

KAJ wrote:
Sun Nov 15, 2020 10:26 pm
jimbob wrote:
Sun Nov 15, 2020 9:03 pm
Unfortunately there's a huge lag in that data:

Image
I don't think there is in the way you suggest. The data reported today comes from specimens taken one or more days earlier. For example, here's the head of today's data set as I've downloaded it:
Screenshot 2020-11-15 at 21.41.07.png
There were 24,962 cases reported today (PubCases) but none are shown for today under SpecCases (by specimen date). That doesn't mean those 24,962 cases are missing from SpecCases, they're just shown with the specimen date, not the date of publication. The missing values under SpecCases for the last few days will be filled in when those results become available, at the same time as they are reported. For a given date the numbers in PubCases will not change, the numbers in SpecCases will - but this reflects delays common to both columns but "hidden" in PubCases.

There is a not-very-predictable delay between specimen date and reporting date, but that delay applies equally to both columns of data. Using PubCases ignores that delay, using SpecCases corrects for it.

The latest few days in SpecCases tend to have silly low values, many results from specimens taken on those days are not yet available. When results are available they appear in (different rows of) the SpecCases and PubCases columns at the same time. Because recent SpecCases are probably incomplete, I tend to exclude the latest 3 or 4 days from my analyses.

Associating any datum with the reporting date builds in and ignores the not-very-predictable delay between specimen date and reporting date, this is time-axis noise whose effect is very difficult to reduce. Associating the datum with specimen date means the delay is explicit in the missing data for recent days. As shown in my recent post this allows one to fit and study time effects with very simple models and little unfitted noise.
Ah, we're talking about two different datasets.

I thought you were talking about deaths recorded by date of death, which does indeed have a lag - of over a week.

That isn't the case for deaths by date reported, which are obviously those reported between 4pm and 4pm that day and are fixed.

And obviously you are talking about cases.

:oops:

So we're in agreement at least
Have you considered stupidity as an explanation

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 10:38 pm

lpm wrote:
Sun Nov 15, 2020 10:12 pm
My suspicion is that the UK has:

- hotspot areas that were in tier 3, and to a lesser extent tier 2, where cases have fallen very successfully - more compliance, in place longer

- non hotspot areas that were tier 1 where cases are rising - poor compliance, not been locked down for long

My own area had grown from very low to what would be tier 2 over the last month.

It doesn't look like 2 more weeks of lockdown is anything like enough to get Covid knocked back. R=1 doesn't get us anywhere and there's no sign R is at 0.7 nationally.
That's my thinking too.
Have you considered stupidity as an explanation

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 10:41 pm

bob sterman wrote:
Sun Nov 15, 2020 7:05 pm
sTeamTraen wrote:
Sun Nov 15, 2020 5:49 pm
Sue Cook, former presenter of Crimewatch, has gone full frootloop.
She appears to have fallen under the spell of Sikora, Heneghan and that diet guru bloke off Youtube...

Ex-BBC star Sue Cook, 71, accuses broadcaster of unbalanced Covid coverage and pushing too hard to shut down UK by forever 'wheeling out' 'Professor Lockdown' Neil Ferguson over other scientists
https://www.dailymail.co.uk/news/articl ... erage.html

The diet guru bloke (Cummins) is described as "Chief Program Officer for Irish Heart Disease Awareness" - which sounds very impressive. Except Irish Heart Disease Awareness has the same address as the one Cummins sells his diet guru books from.
Cummins genuinely seems to believe his statements. Which is surprising.

He also was saying that we need to wait for the end of the year to see if this half year is just going to be like a bad flu. I pointed out that in Europe, you'd need 20,000 negative excess deaths up to Christmas in order to get to the second worst year on the Euromono dataset. It's currently over 50000 and rising. And indeed accelerating.
Have you considered stupidity as an explanation

KAJ
Fuzzable
Posts: 310
Joined: Thu Nov 14, 2019 5:05 pm
Location: UK

Re: COVID-19

Post by KAJ » Sun Nov 15, 2020 11:18 pm

jimbob wrote:
Sun Nov 15, 2020 10:36 pm


Ah, we're talking about two different datasets.

I thought you were talking about deaths recorded by date of death, which does indeed have a lag - of over a week.

That isn't the case for deaths by date reported, which are obviously those reported between 4pm and 4pm that day and are fixed.

And obviously you are talking about cases.

:oops:

So we're in agreement at least
I'm afraid not. I was not clear that although I said "cases" the same argument applied mutatis mutandis to deaths. gov.uk says (my emphasis):
Total number of people who had a positive test result for COVID-19 and died within 28 days of the first positive test, reported on or up to the date of death or reporting date (depending on availability).

People who died more than 28 days after their first positive test are not included, whether or not COVID-19 was the cause of death.

Death data can be presented by when death occurred (date of death) or when the death was reported (date reported) and the availability of each of these time series varies by area:
  • Deaths by date of death - each death is assigned to the date that the person died irrespective of how long it took for the death to be reported. Previously reported data are therefore continually updated
  • Deaths by date reported - each death is assigned to the date when it was first included in the published totals.
There are delays between deaths occurring and their being published in the gov.uk data set. Those delays are not-very-predictable and, as you've noticed, can be quite considerable. To associate deaths with date of publication builds this time-axis noise in and ignores it. Analysing with respect to date of death makes obvious that data for very recent dates is incomplete, but this is how it is.

Compare the charts of deaths by date of death and by date of publication:
Screenshot 2020-11-15 at 23.12.09.png
Screenshot 2020-11-15 at 23.12.09.png (16.39 KiB) Viewed 2729 times
See how using publication date introduces a substantial weekday effect almost absent from the by date of death graph. That size of real weekday effect on deaths just isn't credible - it's caused by reporting delays. For cases, on the other hand, the weekday effect reflects real variations in testing numbers.
chart.png
chart.png (5.44 KiB) Viewed 2729 times

badger
Fuzzable
Posts: 264
Joined: Tue Nov 12, 2019 3:23 pm

Re: COVID-19

Post by badger » Sun Nov 15, 2020 11:26 pm

jimbob wrote:
Sun Nov 15, 2020 10:38 pm
lpm wrote:
Sun Nov 15, 2020 10:12 pm
My suspicion is that the UK has:

- hotspot areas that were in tier 3, and to a lesser extent tier 2, where cases have fallen very successfully - more compliance, in place longer

- non hotspot areas that were tier 1 where cases are rising - poor compliance, not been locked down for long

My own area had grown from very low to what would be tier 2 over the last month.

It doesn't look like 2 more weeks of lockdown is anything like enough to get Covid knocked back. R=1 doesn't get us anywhere and there's no sign R is at 0.7 nationally.
That's my thinking too.
@jamesannan has R=1 now. He has been on the money so far? But yes, can't see restrictions being lifted for 3rd Dec.

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Sun Nov 15, 2020 11:34 pm

KAJ wrote:
Sun Nov 15, 2020 11:18 pm
jimbob wrote:
Sun Nov 15, 2020 10:36 pm


Ah, we're talking about two different datasets.

I thought you were talking about deaths recorded by date of death, which does indeed have a lag - of over a week.

That isn't the case for deaths by date reported, which are obviously those reported between 4pm and 4pm that day and are fixed.

And obviously you are talking about cases.

:oops:

So we're in agreement at least
I'm afraid not. I was not clear that although I said "cases" the same argument applied mutatis mutandis to deaths. gov.uk says (my emphasis):
Total number of people who had a positive test result for COVID-19 and died within 28 days of the first positive test, reported on or up to the date of death or reporting date (depending on availability).

People who died more than 28 days after their first positive test are not included, whether or not COVID-19 was the cause of death.

Death data can be presented by when death occurred (date of death) or when the death was reported (date reported) and the availability of each of these time series varies by area:
  • Deaths by date of death - each death is assigned to the date that the person died irrespective of how long it took for the death to be reported. Previously reported data are therefore continually updated
  • Deaths by date reported - each death is assigned to the date when it was first included in the published totals.
There are delays between deaths occurring and their being published in the gov.uk data set. Those delays are not-very-predictable and, as you've noticed, can be quite considerable. To associate deaths with date of publication builds this time-axis noise in and ignores it. Analysing with respect to date of death makes obvious that data for very recent dates is incomplete, but this is how it is.

Compare the charts of deaths by date of death and by date of publication:
Screenshot 2020-11-15 at 23.12.09.png
See how using publication date introduces a substantial weekday effect almost absent from the by date of death graph. That size of real weekday effect on deaths just isn't credible - it's caused by reporting delays. For cases, on the other hand, the weekday effect reflects real variations in testing numbers.
chart.png
Yes, and you can do a 7day average on the reported date and smooth out most of the noise - yes the maximum lag might be longer than a week, but the biggest variation is in day of the week.

Unsurprisingly, when you do that, you see that when rising, the reported date lags the actual date equivalent rolling average
Have you considered stupidity as an explanation

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Mon Nov 16, 2020 7:34 am

jimbob wrote:
Sun Nov 15, 2020 11:34 pm
KAJ wrote:
Sun Nov 15, 2020 11:18 pm
jimbob wrote:
Sun Nov 15, 2020 10:36 pm


Ah, we're talking about two different datasets.

I thought you were talking about deaths recorded by date of death, which does indeed have a lag - of over a week.

That isn't the case for deaths by date reported, which are obviously those reported between 4pm and 4pm that day and are fixed.

And obviously you are talking about cases.

:oops:

So we're in agreement at least
I'm afraid not. I was not clear that although I said "cases" the same argument applied mutatis mutandis to deaths. gov.uk says (my emphasis):
Total number of people who had a positive test result for COVID-19 and died within 28 days of the first positive test, reported on or up to the date of death or reporting date (depending on availability).

People who died more than 28 days after their first positive test are not included, whether or not COVID-19 was the cause of death.

Death data can be presented by when death occurred (date of death) or when the death was reported (date reported) and the availability of each of these time series varies by area:
  • Deaths by date of death - each death is assigned to the date that the person died irrespective of how long it took for the death to be reported. Previously reported data are therefore continually updated
  • Deaths by date reported - each death is assigned to the date when it was first included in the published totals.
There are delays between deaths occurring and their being published in the gov.uk data set. Those delays are not-very-predictable and, as you've noticed, can be quite considerable. To associate deaths with date of publication builds this time-axis noise in and ignores it. Analysing with respect to date of death makes obvious that data for very recent dates is incomplete, but this is how it is.

Compare the charts of deaths by date of death and by date of publication:
Screenshot 2020-11-15 at 23.12.09.png
See how using publication date introduces a substantial weekday effect almost absent from the by date of death graph. That size of real weekday effect on deaths just isn't credible - it's caused by reporting delays. For cases, on the other hand, the weekday effect reflects real variations in testing numbers.
chart.png
Yes, and you can do a 7day average on the reported date and smooth out most of the noise - yes the maximum lag might be longer than a week, but the biggest variation is in day of the week.

Unsurprisingly, when you do that, you see that when rising, the reported date lags the actual date equivalent rolling average
Screenshot 2020-11-16 073234.png
Screenshot 2020-11-16 073234.png (52.43 KiB) Viewed 2691 times
Have you considered stupidity as an explanation

User avatar
jimbob
Light of Blast
Posts: 5276
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Mon Nov 16, 2020 8:13 am

Though of course, I guess looking at the variation, it would make more sense for it to be a geometric mean rather than a straight average. But I suspect it won't change the story much.

As I'm not discussing this in public with disingenuous fools, I can put it back into a log scale
Screenshot 2020-11-16 081237.png
Screenshot 2020-11-16 081237.png (68.26 KiB) Viewed 2681 times
Have you considered stupidity as an explanation

KAJ
Fuzzable
Posts: 310
Joined: Thu Nov 14, 2019 5:05 pm
Location: UK

Re: COVID-19

Post by KAJ » Mon Nov 16, 2020 9:22 am

jimbob wrote:
Sun Nov 15, 2020 11:34 pm
Yes, and you can do a 7day average on the reported date and smooth out most of the noise - yes the maximum lag might be longer than a week, but the biggest variation is in day of the week.

Unsurprisingly, when you do that, you see that when rising, the reported date lags the actual date equivalent rolling average
But after distorting the data by associating with the wrong (for almost all purposes) date, that introduces a further, artificial, distortion. It may look better, but it's still wrong!

The blunt fact is that (at 5:00 p.m. 15 Nov) we don't know how many cases (or deaths, admissions, inpatients) there were on 15th, 14th, 13th November. Our information for a few days earlier is suspect. Plotting earlier cases (or deaths) on those dates doesn't really help at all, it is very likely to mislead others and perhaps even ourselves. The only honest way of plotting data for 15th, 14th, 13th is to wait until that data is available.

Locked