Coronavirus - Data Analysis Thread

[Smollet]

havoc said:

vaud said:

Terminator X said:

^^ not spreading in the pubs though, our local has had not a single case of covid the entire time.

TX.

Not sure you be so assured given it can be asymptomatic? Plus those who get mild symptoms, write it off as a cold and never get tested.

...and lead-times for infection are typically 5-10 days, with the pubs only open a week.

Yes. The figures come the first May Bank holiday should be interesting.

[Terminator X]

Smollet said:

havoc said:

vaud said:

Terminator X said:

^^ not spreading in the pubs though, our local has had not a single case of covid the entire time.

TX.

Not sure you be so assured given it can be asymptomatic? Plus those who get mild symptoms, write it off as a cold and never get tested.

...and lead-times for infection are typically 5-10 days, with the pubs only open a week.

Yes. The figures come the first May Bank holiday should be interesting.

I'd be amazed if there is a spike, the pubs are not places to catch covid.

As for my comment "has had not a single case of covid the entire time" I mean nothing has come up for them over the last 12-13 months using track / trace / covid app / barcode scans / whatever.

TX.

[havoc]

First point, I have to disagree - crowded indoor space? Fits the criteria.

Second - I can't say I'm surprised...track and trace has been a ridiculously fking expensive farce which has done little but line the pockets of Harding and her mates...

[Elysium]

Quick update on todays data. We are a full week after retail and beer gardens reopened, so we should start to see signs of an uptick in cases if this has increased transmission. I also wanted to see how many people were testing following the offer of twice weekly LFD for us all:

1. Tests and cases. Tests have increased, but not by as much as we might have expected. Average of 936/day over the last week. Despite LFD cases are falling back in line with the 18 day halving we have seen since New Year:



2. Key metrics. Cases per 100k tests and admissions are 'nailed on' the 18 day halving line. Deaths are still halving every 14 days. Big news is that we had just 1 death within 28 days of a positive test in all of the UK on the 18th April. There is probably reporting lag, but its still quite remarkable:







3. Key metrics overlaid. All three are falling predictably and consistently despite each step we have taken to lift restrictions.

[RSTurboPaul]

Elysium said:

Quick update on todays data. We are a full week after retail and beer gardens reopened, so we should start to see signs of an uptick in cases if this has increased transmission. I also wanted to see how many people were testing following the offer of twice weekly LFD for us all:

1. Tests and cases. Tests have increased, but not by as much as we might have expected. Average of 936/day over the last week. Despite LFD cases are falling back in line with the 18 day halving we have seen since New Year:



2. Key metrics. Cases per 100k tests and admissions are 'nailed on' the 18 day halving line. Deaths are still halving every 14 days. Big news is that we had just 1 death within 28 days of a positive test in all of the UK on the 18th April. There is probably reporting lag, but its still quite remarkable:







3. Key metrics overlaid. All three are falling predictably and consistently despite each step we have taken to lift restrictions.

You really need to send these to JHB et al if you haven't already, Elysium.

It is patently clear that everything the Government claims has influenced the decline, er, hasn't...

[Elysium]

Update on ONS data released today. No excess deaths again in week 14 (ending 9th April) and deaths due to COVID are counted at 280

6 consecutive weeks of normal deaths. The last 5 at or below the 5 year minimum.

[768]

Elysium said:

Quick update on todays data. We are a full week after retail and beer gardens reopened, so we should start to see signs of an uptick in cases if this has increased transmission.

Should start as in, should be able to see now or should be able to see in future datasets from now? I assume it's still thought it takes around a week for most infections to be detectable and then you need to accrue some data to discern a trend? So, give it a couple more weeks and we might have a signal?

[WatchfulEye]

Elysium said:

Quick update on todays data. We are a full week after retail and beer gardens reopened, so we should start to see signs of an uptick in cases if this has increased transmission. I also wanted to see how many people were testing following the offer of twice weekly LFD for us all:

A lengthening of the halving time is also indicative of increased transmission. So, if the halving time increases above the 18 days which seems to be fairly consistent over the last few weeks, then we will have evidence of increased transmission.

[WindyCommon]

WatchfulEye said:

A lengthening of the halving time is also indicative of increased transmission. So, if the halving time increases above the 18 days which seems to be fairly consistent over the last few weeks, then we will have evidence of increased transmission.

I am not sure that is true - perhaps more accurately relevant - once prevalence drops below a certain threshold. Similarly Rt is isn’t useful below a certain prevalence threshold. We have to keep our eyes on the absolute number of cases as well as the changes.

[havoc]

Interesting article on probabilities, false-positives and Bayesian mathematics:-

https://www.theguardian.com/world/2021/apr/18/obsc...

[Terminator X]

havoc said:

Interesting article on probabilities, false-positives and Bayesian mathematics:-

https://www.theguardian.com/world/2021/apr/18/obsc...

Didn't quite understand it all other than it seems a tad unwise to be testing so many people with no symptoms given a "reasonable amount"^ may think they are positive when they actually aren't?

^or the Govt think it of course.

TX.

[havoc]

False positives in this context aren't really an issue:-
- As mentioned, they'll then get PCR tests which SHOULD make it certain or not.
- A false positive isolating has no risk to anyone else.


The issue here (as with most medical tests) is false negatives...which the article unfortunately doesn't explore enough. You can infer the type and size of the issue though by substituting positive for negative.

[vaud]

As an aside from UK analysis, I am watching the situation closely in India as we have many staff there.

Truly terrifying data:

Projecting 14,000 deaths / day by May 17 / 1 million by August
259k beds available, 3.3million beds needed at peak

https://covid19.healthdata.org/india?view=daily-de...

Any other (good) data sources for India?

[vaud]

anonymous said:

[redacted]

Thanks Tonker. My wording was indeed ambiguous as to "by Aug 1" and should have read "total of 1 million by Aug 1". An error on my part which I am happy to correct.

Now, to the point of stats, as I was trying to articulate that I am looking for more than one source, do you have any?

I know you are very angry on this topic.

[Elysium]

Update on todays data:

1. Tests and Cases. The increase in testing has been pretty slow since the Govt offered weekly LFD to the entire nation. Despite the testing volumes cases are regressing the 18 day halving line.



2. Key metrics. Cases per 100k tests and admissions continue to halve every 18 days. I have added a CFR line to the cases graph, which has fallen from over 3% at the peak to an average of 0.59% over the last 2 weeks.





Deaths are still halving every 14 days but dont really have anywere further to go. Averaging 12 deaths per day in the last week and 19 deaths per day in the week before.




3. This shows all three metrics overlaid. The second graph zooms in on the last 4 months of steady decline.

[Elysium]

vaud said:

As an aside from UK analysis, I am watching the situation closely in India as we have many staff there.

Truly terrifying data:

Projecting 14,000 deaths / day by May 17 / 1 million by August
259k beds available, 3.3million beds needed at peak

https://covid19.healthdata.org/india?view=daily-de...

Any other (good) data sources for India?

Our World in Data is useful. Unfortunately, on a per capita basis, India is only around 25% of the way to our winter peak. Their healthcare capacity is clearly even more limited than our own.

https://ourworldindata.org/explorers/coronavirus-d...

[Jambo85]

Great update as usual Elysium, encouraging too (for UK).

What conclusions do you draw from the CFR rate? Optimistically the recent decline could be vaccine related but I note it has been at low levels before also - is that a function of healthcare overloading perhaps?

[vaud]

Elysium said:

Our World in Data is useful. Unfortunately, on a per capita basis, India is only around 25% of the way to our winter peak. Their healthcare capacity is clearly even more limited than our own.

Thanks. I understand testing capacity is limited in India as well... And political manipulation. On the ground data (as I have heard) indicates that the govt is underreporting (e.g. Govt reports 10 deaths, local funeral pyres and queues indicate far, far more)

[vaud]

Thank you for taking the time to construct a detailed reply.

My "on the ground intel" is that it is very bad at the moment and escalating quickly. Our company has >100k workforce in India, and while mostly <50 years old it is spreading like wildfire, which will impact the Indian economy as well as the increased mortality.

[Jambo85]

anonymous said:

[redacted]

It's hard to trade various things off against each other but while you have listed things undoubtedly in India's favour, there are a fair few things working the other way:
- Lower standard/quantity of healthcare
- Vast numbers of people living in poverty with presumably less access to healthcare
- Those in poverty presumably less educated and with less access to reliable information and sanitation to protect themselves and others
- It is common in India to have very large households of not just multiple generations but multiple families.
- Vast, densely populated cities
ETA:
- Poverty = living hand to mouth - so going out to get food daily or more, unlike us lucky people who can buy enough for a week or more or even have it delivered.
- Also the same people if they don't work that day they don't eat that day - I don't know for sure but I'm guessing the state isn't paying people 80% of their wage to sit at home and do feck all - something which likely helped here.

Edited by Jambo85 on Monday 26th April 23:29