COVID-19

Covid-19 discussion, bring your own statistics
Post Reply
User avatar
Little waster
Catbabel
Posts: 893
Joined: Tue Nov 12, 2019 12:35 am
Location: About 1 inch behind my eyes

Re: COVID-19

Post by Little waster » Thu Oct 08, 2020 10:47 am

jimbob wrote:
Thu Oct 08, 2020 10:29 am
we know that superspreading events are highly significant.
I know Trump rallies are unpleasant but I doubt they'll have any lasting effect historically.
People who use Godwin's Law are worse than Hitler.

User avatar
lpm
After Pie
Posts: 1848
Joined: Mon Nov 11, 2019 1:05 pm
Location: INJECT HIM WITH BLEACH

Re: COVID-19

Post by lpm » Thu Oct 08, 2020 11:05 am

jimbob wrote:
Thu Oct 08, 2020 10:29 am
Exactly - talking about R gives the illusion of knowing more than we do about this epidemic. And it's not going to be useful in a proper model as we know that superspreading events are highly significant.
No no no no no. The exact opposite. It's the "proper models" that have given the illusion of knowing more than we do. It's R that gives the clarity.

Why do we need a proper model, for practical action terms?

All we need to know is currently R>1. Significantly so.

Therefore LockdownLevel(8Oct) < LockdownLevel(R=1). Significantly so.

Therefore we need to increase LockdownLevel significantly.

This is like planing a diet, and spending weeks coming up with a model that includes the impacts of extra exercise and lower cake to calculate the expected weight loss per week, but not actual starting the diet till this is done. Just go for a run and eat less cake, then see how it goes. Keep taking action till your weight loss demonstrates the calorie ratio is <1.

This government has always had an illusion that fine-tuning is possible with this accursed virus. Keeps running its little models, which are always wrong, and then thinking it can dial things up and down. It's truly remarkable that they thought they could tweak the lockdown level with tinkering little actions like shutting pubs a bit early. The governments idea that they'd move little local areas in and out of measures has been proved to be a disaster - simply because the underlying national lockdown level is far too low and the infections constantly bubble up again.

Local lockdowns are like saying the diet overall is fine, our model shows it's just the pasta eating that's in excess so we'll tackle the pasta but carry on with the cake for breakfast. Stop eating the f.cking cake as well.
I'll miss him after he's died in the pandemic

User avatar
Brightonian
Snowbonk
Posts: 381
Joined: Mon Nov 11, 2019 3:16 pm

Re: COVID-19

Post by Brightonian » Thu Oct 08, 2020 11:56 am

Actually, what I would personally like is the percentage of non-quarantining infectious people in my area so I can judge how likely I will come across them. There are some graphs like Worldometers' "Active Cases" (e.g. a third of the way down this page), but there's no info on how many of the cases are known about (and hence quarantining (unless you're Trump)).

User avatar
shpalman
After Pie
Posts: 1991
Joined: Mon Nov 11, 2019 12:53 pm
Location: One step beyond

Re: COVID-19

Post by shpalman » Thu Oct 08, 2020 12:18 pm

lpm wrote:
Thu Oct 08, 2020 11:05 am
jimbob wrote:
Thu Oct 08, 2020 10:29 am
Exactly - talking about R gives the illusion of knowing more than we do about this epidemic. And it's not going to be useful in a proper model as we know that superspreading events are highly significant.
No no no no no. The exact opposite. It's the "proper models" that have given the illusion of knowing more than we do. It's R that gives the clarity.

Why do we need a proper model, for practical action terms?

All we need to know is currently cases are going up. Significantly so.

Therefore LockdownLevel(8Oct) < LockdownLevel(cases not going up). Significantly so.

Therefore we need to increase LockdownLevel significantly...
molto tricky

badger
Clardic Fug
Posts: 154
Joined: Tue Nov 12, 2019 3:23 pm

Re: COVID-19

Post by badger » Thu Oct 08, 2020 1:39 pm

Meanwhile...

Eye catching headline: outbreak Covid: more than 80% of positive UK cases in study had no core symptoms

Some detail:
The study, reported in Clinical Epidemiology, analysed the symptoms described by more than 36,000 people who were tested between April and June. Only 115 tests came back positive and of those only 27 people, or 23.5%, had symptoms of any description.
Actual paper:

https://www.dovepress.com/three-quarter ... ticle-CLEP

Does this chime with what is known already in terms of percentage asymptomatic population? And is there anything other than guesswork about transmission by asymptomatics (rather than presymptomatics)?

User avatar
lpm
After Pie
Posts: 1848
Joined: Mon Nov 11, 2019 1:05 pm
Location: INJECT HIM WITH BLEACH

Re: COVID-19

Post by lpm » Thu Oct 08, 2020 3:15 pm

17,540.

Soon we'll be beating all the EU but for now we're still a bit behind France.
I'll miss him after he's died in the pandemic

User avatar
jimbob
After Pie
Posts: 1621
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Thu Oct 08, 2020 4:30 pm

The number of coronavirus patients in intensive care in the north of England will surpass the April peak if infections continue rising at the current rate, MPs have been warned in a briefing chaired by Chris Whitty and a minister.

MPs were also shown early research by Public Health England suggesting that bars, pubs and restaurants accounted for 41% of cases in which two or more under-30s had visited the same venue in the week before testing positive. This fell to a quarter of infections across all age groups, the MPs were told.
https://www.theguardian.com/world/2020/ ... in-22-days

Eat out to Help Covid Out
Have you considered stupidity as an explanation

User avatar
jdc
Hilda Ogden
Posts: 549
Joined: Wed Sep 25, 2019 4:31 pm

Re: COVID-19

Post by jdc » Thu Oct 08, 2020 6:49 pm

badger wrote:
Thu Oct 08, 2020 1:39 pm
Meanwhile...

Eye catching headline: outbreak Covid: more than 80% of positive UK cases in study had no core symptoms

Some detail:
The study, reported in Clinical Epidemiology, analysed the symptoms described by more than 36,000 people who were tested between April and June. Only 115 tests came back positive and of those only 27 people, or 23.5%, had symptoms of any description.
Actual paper:

https://www.dovepress.com/three-quarter ... ticle-CLEP

Does this chime with what is known already in terms of percentage asymptomatic population? And is there anything other than guesswork about transmission by asymptomatics (rather than presymptomatics)?
Can't remember which thread now but I posted a paper that followed up 'asymptomatics'. It was a tiny number of people in the study, mind. iirc 10 of the 13 asymptos were actually presymptomatic and developed symptoms within ?7 days.

headline should end "on the day of the test"

User avatar
discovolante
After Pie
Posts: 1886
Joined: Fri Oct 11, 2019 5:10 pm

Re: COVID-19

Post by discovolante » Thu Oct 08, 2020 6:54 pm

jdc wrote:
Thu Oct 08, 2020 6:49 pm
badger wrote:
Thu Oct 08, 2020 1:39 pm
Meanwhile...

Eye catching headline: outbreak Covid: more than 80% of positive UK cases in study had no core symptoms

Some detail:
The study, reported in Clinical Epidemiology, analysed the symptoms described by more than 36,000 people who were tested between April and June. Only 115 tests came back positive and of those only 27 people, or 23.5%, had symptoms of any description.
Actual paper:

https://www.dovepress.com/three-quarter ... ticle-CLEP

Does this chime with what is known already in terms of percentage asymptomatic population? And is there anything other than guesswork about transmission by asymptomatics (rather than presymptomatics)?
Can't remember which thread now but I posted a paper that followed up 'asymptomatics'. It was a tiny number of people in the study, mind. iirc 10 of the 13 asymptos were actually presymptomatic and developed symptoms within ?7 days.

headline should end "on the day of the test"
Ugh, seen stuff going around with people commenting about number of 'symptomless' cases meaning Covid isn't a risk etc etc, not great eh.
don't get any big ideas, they're not gonna happen

badger
Clardic Fug
Posts: 154
Joined: Tue Nov 12, 2019 3:23 pm

Re: COVID-19

Post by badger » Thu Oct 08, 2020 7:44 pm

jdc wrote:
Thu Oct 08, 2020 6:49 pm
badger wrote:
Thu Oct 08, 2020 1:39 pm
Meanwhile...

Eye catching headline: outbreak Covid: more than 80% of positive UK cases in study had no core symptoms

Some detail:
The study, reported in Clinical Epidemiology, analysed the symptoms described by more than 36,000 people who were tested between April and June. Only 115 tests came back positive and of those only 27 people, or 23.5%, had symptoms of any description.
Actual paper:

https://www.dovepress.com/three-quarter ... ticle-CLEP

Does this chime with what is known already in terms of percentage asymptomatic population? And is there anything other than guesswork about transmission by asymptomatics (rather than presymptomatics)?
Can't remember which thread now but I posted a paper that followed up 'asymptomatics'. It was a tiny number of people in the study, mind. iirc 10 of the 13 asymptos were actually presymptomatic and developed symptoms within ?7 days.

headline should end "on the day of the test"
Yes it should and I can't believe that it's not in the article or in the paper (as much as I've scanned). I had misread it on the assumption that if it's just on the day of the test that it doesn't tell us much, so why bother reporting it? Thankfully (for me, but probably not for anyone else!) I posted on here as my critical faculties were clearly on the blink.

User avatar
sTeamTraen
Catbabel
Posts: 930
Joined: Mon Nov 11, 2019 4:24 pm

Re: COVID-19

Post by sTeamTraen » Thu Oct 08, 2020 8:48 pm

badger wrote:
Thu Oct 08, 2020 7:44 pm
jdc wrote:
Thu Oct 08, 2020 6:49 pm
Can't remember which thread now but I posted a paper that followed up 'asymptomatics'. It was a tiny number of people in the study, mind. iirc 10 of the 13 asymptos were actually presymptomatic and developed symptoms within ?7 days.

headline should end "on the day of the test"
Yes it should and I can't believe that it's not in the article or in the paper (as much as I've scanned). I had misread it on the assumption that if it's just on the day of the test that it doesn't tell us much, so why bother reporting it? Thankfully (for me, but probably not for anyone else!) I posted on here as my critical faculties were clearly on the blink.
The phrase "on the day of the test" appears four times in the Results paragraph of the Abstract of the paper, and the Conclusions paragraph focuses on the fact that "Having no symptoms doesn't mean no infection". This seems quite responsible to me, so the spin is probably either at the level of the UCL press office or the media outlets.
Something something hammer something something nail

User avatar
lpm
After Pie
Posts: 1848
Joined: Mon Nov 11, 2019 1:05 pm
Location: INJECT HIM WITH BLEACH

Re: COVID-19

Post by lpm » Fri Oct 09, 2020 12:51 pm

Based on the current curve, a quick back-of-envelope calc gives 200 to 1,000 extra UK deaths for every day the government delays action.

We've made this mistake before. Delay in the hope the numbers magically turn downwards.

It'll be interesting to see if the public acts before the govt. We did in March, shutting down well ahead of government commandments. I suspect we won't so much this time, our civic responsibility has been totally Cummingsed. But I bet this weekend will still see a drop off nonetheless.
I'll miss him after he's died in the pandemic

User avatar
Bird on a Fire
Stummy Beige
Posts: 3804
Joined: Fri Oct 11, 2019 5:05 pm
Location: nadir of brie

Re: COVID-19

Post by Bird on a Fire » Fri Oct 09, 2020 1:13 pm

If the current spikes were caused by students moving long-distance and mingling, as long as university populations stay pretty static those new clusters should burn out in a month or two.

Luckily there's nothing happening in mid-December that might cause people to move long-distance and mingle.
“One of the penalties of an ecological education is that one lives alone in a world of wounds.”
― Aldo Leopold, A Sand County Almanac (1949)

#ShowYourStripes

User avatar
lpm
After Pie
Posts: 1848
Joined: Mon Nov 11, 2019 1:05 pm
Location: INJECT HIM WITH BLEACH

Re: COVID-19

Post by lpm » Fri Oct 09, 2020 1:15 pm

Why would clusters burn out? Are you suggesting students reach herd immunity? There's plenty more campuses to get outbreaks.
I'll miss him after he's died in the pandemic

User avatar
jimbob
After Pie
Posts: 1621
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Fri Oct 09, 2020 1:29 pm

Just catching up on the In the pipeline blog and this was interesting on Interferon and especially those wit antibodies to interferons

https://blogs.sciencemag.org/pipeline/a ... oronavirus
In this latest work, the authors look for the intersection of these stories: pre-existing autoantibodies in coronavirus patients that neutralize one or more types of interferon. The results are striking: in 1,227 healthy controls, four patients were found to have such antibodies. In 663 patients with mild coronavirus infections, none were found at all. But in 987 severely ill patients, at least 101 of them had such antibodies against at least one of the Type I interferons (!) I hope that those four people identified in the first cohort are taking precautions, was my first thought, because it certainly looks like they in a newly identified risk group. Those numbers come out to p-values of less than ten to the minus sixteenth, which means that we all pretty much have to stand up take take off our hats – this one’s real. The anti-interferon antibody patients also skewed older, which suggests that the levels go up over the years (exactly what you don’t want, honestly).
Have you considered stupidity as an explanation

User avatar
Bird on a Fire
Stummy Beige
Posts: 3804
Joined: Fri Oct 11, 2019 5:05 pm
Location: nadir of brie

Re: COVID-19

Post by Bird on a Fire » Fri Oct 09, 2020 2:19 pm

lpm wrote:
Fri Oct 09, 2020 1:15 pm
Why would clusters burn out? Are you suggesting students reach herd immunity? There's plenty more campuses to get outbreaks.
I'm assuming they'll lock down campuses properly again, which makes isolating infectious clusters a bit easier.
“One of the penalties of an ecological education is that one lives alone in a world of wounds.”
― Aldo Leopold, A Sand County Almanac (1949)

#ShowYourStripes

User avatar
Sciolus
Fuzzable
Posts: 338
Joined: Mon Nov 11, 2019 6:42 pm

Re: COVID-19

Post by Sciolus » Fri Oct 09, 2020 3:25 pm

Bird on a Fire wrote:
Fri Oct 09, 2020 1:13 pm
Luckily there's nothing happening in mid-December that might cause people to move long-distance and mingle.
And it's not like there was anything in China in late January that might have done the same.

User avatar
Woodchopper
After Pie
Posts: 1869
Joined: Sat Oct 12, 2019 9:05 am

Re: COVID-19

Post by Woodchopper » Fri Oct 09, 2020 3:46 pm

lpm wrote:
Fri Oct 09, 2020 1:15 pm
Why would clusters burn out? Are you suggesting students reach herd immunity? There's plenty more campuses to get outbreaks.

Indeed, or local people when the students buy food etc

User avatar
bob sterman
Fuzzable
Posts: 275
Joined: Mon Nov 11, 2019 10:25 pm

Re: COVID-19

Post by bob sterman » Sat Oct 10, 2020 6:56 am

CEBM have been a bit quiet recently but they're back - just in the nick of time with a new message...

Yes - rise in cases is real - but don't lockdown community activities (e.g. pubs, restaurants) etc because (it is claimed) 18% of new COVID-19 hospitalized cases acquired the virus in a healthcare setting (i.e. ONLY 82% of new COVID-19 hospitalized cases appear to be acquired in the community). Given that only about 5% of COVID cases are hospitalised at all - they are really talking about 1% of cases overall.

They way it works is they publish an apparently dispassionate analysis on the CEBM blog...

Probable healthcare associated infections in England
https://www.cebm.net/covid-19/probable- ... n-england/

Then Heneghan adds the political message ("don't shut the pubs!!") by Tweeting about an article which reports on the blog post...

Government urged to 'pause' new restrictions as cases of Covid caught in hospital soar
https://twitter.com/carlheneghan/status ... 6320136194

User avatar
sTeamTraen
Catbabel
Posts: 930
Joined: Mon Nov 11, 2019 4:24 pm

Re: COVID-19

Post by sTeamTraen » Sat Oct 10, 2020 4:18 pm

jimbob wrote:
Fri Oct 09, 2020 1:29 pm
Just catching up on the In the pipeline blog and this was interesting on Interferon and especially those wit antibodies to interferons

https://blogs.sciencemag.org/pipeline/a ... oronavirus
In this latest work, the authors look for the intersection of these stories: pre-existing autoantibodies in coronavirus patients that neutralize one or more types of interferon. The results are striking: in 1,227 healthy controls, four patients were found to have such antibodies. In 663 patients with mild coronavirus infections, none were found at all. But in 987 severely ill patients, at least 101 of them had such antibodies against at least one of the Type I interferons (!) I hope that those four people identified in the first cohort are taking precautions, was my first thought, because it certainly looks like they in a newly identified risk group. Those numbers come out to p-values of less than ten to the minus sixteenth, which means that we all pretty much have to stand up take take off our hats – this one’s real. The anti-interferon antibody patients also skewed older, which suggests that the levels go up over the years (exactly what you don’t want, honestly).
That p value is presumably for the 3x2 chi square, and you can see from a simple visual inspection that it's going to be homeopathic. (The 10^-16 thing just means that the software gave up; my guess is that it's a lot smaller than that.) But you still have 876 out of 977 severely ill patients without these antibodies. So it would seem to be explaining only 10-12% of the variance in outcome (which, granted, is still nice to have), and we don't know what other interactions are going on.

But I bet that when Toby Young and co get hold of this, we'll hear "Test people for this and lock up the ones who have high levels, so I can go to the pub".
Something something hammer something something nail

User avatar
JQH
Catbabel
Posts: 955
Joined: Mon Nov 11, 2019 3:30 pm
Location: Sar Flandan

Re: COVID-19

Post by JQH » Sat Oct 10, 2020 6:59 pm

sTeamTraen wrote:
Sat Oct 10, 2020 4:18 pm


But I bet that when Toby Young and co get hold of this, we'll hear "Test people for this and lock up the ones who have high levels, so I can go to the pub".
Unless they are the ones with high levels; in which case restricting their movements will be a gross infringement of civil liberty.
And remember that if you botch the exit, the carnival of reaction may be coming to a town near you.

Fintan O'Toole

User avatar
sTeamTraen
Catbabel
Posts: 930
Joined: Mon Nov 11, 2019 4:24 pm

Re: COVID-19

Post by sTeamTraen » Tue Oct 13, 2020 4:41 pm

143 deaths reported in the UK today - the highest single-day total since 11 June.

More concerning, though, is the way that the 7-day moving average of deaths is very closely tracking the 7-day moving average of new cases, with a lag that looks like 30-40 days --- far longer than in the first wave. (Image by me with data from ECDC; apologies for the clunky X-axis with arbitrary start point.)

It seems to me that one possible explanation of this is that the schoolkids or young pub drinkers who get infected at T0 pass it on to their parents at T1 and their older relatives at T2, cf. the heatmaps that we've had here recently. (In the first wave that presumably also happened, but we didn't know about it because the T0 and T1 people mostly didn't get tested.) If that continues, we'll be looking at 1000 deaths per day this time next month. :o

UK.png
UK.png (19.04 KiB) Viewed 262 times
Something something hammer something something nail

User avatar
shpalman
After Pie
Posts: 1991
Joined: Mon Nov 11, 2019 12:53 pm
Location: One step beyond

Re: COVID-19

Post by shpalman » Tue Oct 13, 2020 4:45 pm

Well by definition it can't be the same patients dying after 30-40 days, can it? Officially if you survive 28 days you don't die of covid.
molto tricky

User avatar
jimbob
After Pie
Posts: 1621
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Tue Oct 13, 2020 4:50 pm

sTeamTraen wrote:
Tue Oct 13, 2020 4:41 pm
143 deaths reported in the UK today - the highest single-day total since 11 June.

More concerning, though, is the way that the 7-day moving average of deaths is very closely tracking the 7-day moving average of new cases, with a lag that looks like 30-40 days --- far longer than in the first wave. (Image by me with data from ECDC; apologies for the clunky X-axis with arbitrary start point.)

It seems to me that one possible explanation of this is that the schoolkids or young pub drinkers who get infected at T0 pass it on to their parents at T1 and their older relatives at T2, cf. the heatmaps that we've had here recently. (In the first wave that presumably also happened, but we didn't know about it because the T0 and T1 people mostly didn't get tested.) If that continues, we'll be looking at 1000 deaths per day this time next month. :o


UK.png
A quick eyeballing suggests to me that the death rate is climbing faster than the case rate - which suggests the cases are being undercounted and the ratio is getting worse

Data from ECDC - with moving average centred on date. https://www.ecdc.europa.eu/en/publicati ... -worldwide
Screenshot 2020-10-13 175007.png
Screenshot 2020-10-13 175007.png (42.48 KiB) Viewed 255 times
Have you considered stupidity as an explanation

User avatar
nekomatic
Fuzzable
Posts: 290
Joined: Mon Nov 11, 2019 3:04 pm

Re: COVID-19

Post by nekomatic » Tue Oct 13, 2020 11:54 pm

I’m a bit late catching up on this but
Grumble wrote:
Mon Oct 05, 2020 12:19 pm
I rely on excel for all sorts of stuff that a database would be better for, but I can’t get training in databases and we keep going “how much” when we get quotes. So it never gets sorted. I’ve tried to teach myself Access but I can’t get my head round it. It’s like the instructions are written in a foreign language.
Have you got an example you can share of the sort of use case you have in mind? It’s just that the set of problems for which Excel is terrible but Access would be great is rather small tbh.

Hannah Fry came on Lauren Laverne’s 6 Music show last week and described her top five worst mistakes people have made with spreadsheets, in which this current one was number five, and it made me laugh louder than anything I can remember on the radio in a very long time. I’ll find the iPlayer Sounds link.

Post Reply