Coronavirus - Data Analysis Thread
Discussion
Elysium said:
I think 15 million is accepted as a sensible minimum at this point
https://news.sky.com/story/covid-19-15-3-of-englan...Just saw this from a month ago...15.3% of England by mid-Jan (= c.£7m people, so closing on 10m for UK as a whole, allowing for the lower rates in the other nations). Add in 7-8 more weeks and, yeah, I can see how it could hit 15m...but that still seems a little high...based on that headline and the Feb rates I'd be more inclined towards 13m, +/- 10%. Similar ball-park though.
...and I'm still shocked at that level. Maybe I move (or more recently, don't move) in the wrong/right circles. And maybe that explains (both ways) the malaise that many seem to be feeling about the ongoing lockdowns and restrictions.
Edited by havoc on Wednesday 10th March 15:45
Murph7355 said:
If it's not immunity from exposure, then what is causing their curves to mirror ours with no lockdowns and a lower vaccination rate (materially lower?)?? Is another possibility that whilst they haven't had lockdowns, people are just far more observant of Social Distancing and better hygiene?
Sweden have progressively tightened restrictions form January onwards, while not a lockdown in the strictest sense, the restrictions are somewhat similar to the "tier 3" restrictions in the UK. (e.g. pubs/restuarants table service only with occupancy limits and restricted hours, public gatherings banned, amateur sport banned, occupancy limits on shops, etc.).This level of restriction in the UK was sufficient to bring numbers down at the end of last year, and given that spread in Sweden has been generally slower than in the UK (for reasons which are not entirely clear), it's entirely plausible that the drop in numbers in Sweden has been due to the restrictions.
WatchfulEye said:
...given that spread in Sweden has been generally slower than in the UK (for reasons which are not entirely clear)
More rural population on average?More outdoorsy lifestyle?
Fitter lifestyle - less obese / diabetic / hypertensive?
General trust in their government and in 'rules', so arguably a higher % of true adherence to prophylactic measures such as mask-wearing and keeping distance?
Probably others, but that's off the top of my head.
havoc said:
Elysium said:
I think 15 million is accepted as a sensible minimum at this point
https://news.sky.com/story/covid-19-15-3-of-englan...Just saw this from a month ago...15.3% of England by mid-Jan (= c.£7m people, so closing on 10m for UK as a whole, allowing for the lower rates in the other nations). Add in 7-8 more weeks and, yeah, I can see how it could hit 15m...but that still seems a little high...based on that headline and the Feb rates I'd be more inclined towards 13m, +/- 10%. Similar ball-park though.
...and I'm still shocked at that level. Maybe I move (or more recently, don't move) in the wrong/right circles. And maybe that explains (both ways) the malaise that many seem to be feeling about the ongoing lockdowns and restrictions.
At a 1% IFR you would need 12,498,700 infections to result in the 124,987 deaths PHE have recorded within 28 days of a positive test.
At a 0.6% IFR, as per the initial estimates provided by SAGE, you would need 20,831,166 infections
This graph attempts to show the enormous scale of these hidden infections:
The ONS infection survey reported an estimate of 1,122,000 people with Coronavirus at the winter peak:
https://www.ons.gov.uk/peoplepopulationandcommunit...
Elysium said:
You can sense check this working backwards from deaths using an estimate of the overall IFR.
At a 1% IFR you would need 12,498,700 infections to result in the 124,987 deaths PHE have recorded within 28 days of a positive test.
At a 0.6% IFR, as per the initial estimates provided by SAGE, you would need 20,831,166 infections
Makes sense.At a 1% IFR you would need 12,498,700 infections to result in the 124,987 deaths PHE have recorded within 28 days of a positive test.
At a 0.6% IFR, as per the initial estimates provided by SAGE, you would need 20,831,166 infections
Couple of challenges (tell me to shut-up if this shouldn't be on this thread)...
- There has been some, ah, doubt over the acuracy of some deaths which have been certified as Covid. 125k could be overstating matters.
- Surely the IFR is an unknown - it's a 3-way equation - numerator (deaths, known), denominator (infections, unknown), and IFR (rate, calculated or inferred)
- Further, the IFR will vary depending upon the demographic profile of the (infected) population...a tendency towards more minorities* or older people will skew the IFR upwards.
* Who live more usually in cities...and in higher density housing...both of which increase susceptibility to infection.
havoc said:
Elysium said:
You can sense check this working backwards from deaths using an estimate of the overall IFR.
At a 1% IFR you would need 12,498,700 infections to result in the 124,987 deaths PHE have recorded within 28 days of a positive test.
At a 0.6% IFR, as per the initial estimates provided by SAGE, you would need 20,831,166 infections
Makes sense.At a 1% IFR you would need 12,498,700 infections to result in the 124,987 deaths PHE have recorded within 28 days of a positive test.
At a 0.6% IFR, as per the initial estimates provided by SAGE, you would need 20,831,166 infections
Couple of challenges (tell me to shut-up if this shouldn't be on this thread)...
- There has been some, ah, doubt over the acuracy of some deaths which have been certified as Covid. 125k could be overstating matters.
- Surely the IFR is an unknown - it's a 3-way equation - numerator (deaths, known), denominator (infections, unknown), and IFR (rate, calculated or inferred)
- Further, the IFR will vary depending upon the demographic profile of the (infected) population...a tendency towards more minorities* or older people will skew the IFR upwards.
* Who live more usually in cities...and in higher density housing...both of which increase susceptibility to infection.
However, the ONS infection survey includes a regular estimate of the proportion of the population infected with SARS-CoV-2 at a given time. This graph compares that estimate with an estimate of infections, derived from the PHE 'deaths within 28 days' and the following assumptions:
1. 21 days from infection to death
2. An overall IFR of 0.6%.
3. An average 10 day period during which a newly infected person will remain infected
I think the two curves are a decent fit, which supports my belief that the total number of people infected by this virus could easily be around 20 million.
Elysium said:
I think the two curves are a decent fit, which supports my belief that the total number of people infected by this virus could easily be around 20 million.
Final thought from me and I'm off to bed.You may be right, but one thing niggles here...and my concern HAS happened before with data from different sources that appear supportive...
BOTH of those curves are estimates. Is it possible that one curve is using data / estimates from the other set? Or that they're both using the same initial assumptions. (e.g. is the ONS data-set extrapolating using PHE data as part of the formula?)
havoc said:
Elysium said:
I think the two curves are a decent fit, which supports my belief that the total number of people infected by this virus could easily be around 20 million.
Final thought from me and I'm off to bed.You may be right, but one thing niggles here...and my concern HAS happened before with data from different sources that appear supportive...
BOTH of those curves are estimates. Is it possible that one curve is using data / estimates from the other set? Or that they're both using the same initial assumptions. (e.g. is the ONS data-set extrapolating using PHE data as part of the formula?)
https://www.ons.gov.uk/peoplepopulationandcommunit...
This is entirely unrelated to the PHE number which is based on deaths within 2 days of a PCR test.
The total of each bar is the ONS mid 2019 population projection for each age group. The red is the number dying with COVID and the yellow is the number dying of anything else.
This is how this looks like as a percentage of the population in each age range:
The unsurprising conclusion is that very few of us have died with COVID and that the people dying with COVID tend to be the same sort of people that die of other things.
Not healthy people of working age.
This is how this looks like as a percentage of the population in each age range:
The unsurprising conclusion is that very few of us have died with COVID and that the people dying with COVID tend to be the same sort of people that die of other things.
Not healthy people of working age.
Update with todays numbers:
1. Tests and cases. The massive increase in asymptomatic LFD testing of school children has failed to produce a corresponding increase in cases, which are continuing to halve in number every 18 days:
2. Key metrics. Cases per 100k tests and admissions continue to halve every 18 days. Deaths are falling more linearly:
3. All three metrics overlaid and moving roughly together
1. Tests and cases. The massive increase in asymptomatic LFD testing of school children has failed to produce a corresponding increase in cases, which are continuing to halve in number every 18 days:
2. Key metrics. Cases per 100k tests and admissions continue to halve every 18 days. Deaths are falling more linearly:
3. All three metrics overlaid and moving roughly together
The Govt website has been updated to include numbers of LFD and PCR cases for England.
This graph shows daily PCR tests in orange and LFD in blue. The red dotted line is the 7 day average of PCR cases and the barely visible yellow dotted line is the 7 day average of LFD cases.
I was quite shocked when I saw this. LFD really isn't finding cases. In fact, so far we have performed almost 27 million LFD tests, but found a grand total of just 43,106 LFD cases.
For comparison we found 66,799 PCR cases in a single day on the 4th Jan 2021
This next graph compares positivity (cases / tests) for LFD and PCR. They are shown on different y-axis to allow comparison as LFD is much lower. The interesting thing is that they align unexpectedly well:
On average PCR is finding 25 times more cases per test than LFD.
This could be about the difference between symptomatic and asymptomatic prevalence.
Or it could mean PCR is overly sensitive
Or it could mean that LFD is no sensitive enough
Or it could mean that infectious asymptomatic cases are rare.
In any event average LFD positivity for the whole of March, including the school testing is just 0.09%
9 cases per 10,000 tests.
Mind blown.
This graph shows daily PCR tests in orange and LFD in blue. The red dotted line is the 7 day average of PCR cases and the barely visible yellow dotted line is the 7 day average of LFD cases.
I was quite shocked when I saw this. LFD really isn't finding cases. In fact, so far we have performed almost 27 million LFD tests, but found a grand total of just 43,106 LFD cases.
For comparison we found 66,799 PCR cases in a single day on the 4th Jan 2021
This next graph compares positivity (cases / tests) for LFD and PCR. They are shown on different y-axis to allow comparison as LFD is much lower. The interesting thing is that they align unexpectedly well:
On average PCR is finding 25 times more cases per test than LFD.
This could be about the difference between symptomatic and asymptomatic prevalence.
Or it could mean PCR is overly sensitive
Or it could mean that LFD is no sensitive enough
Or it could mean that infectious asymptomatic cases are rare.
In any event average LFD positivity for the whole of March, including the school testing is just 0.09%
9 cases per 10,000 tests.
Mind blown.
Edited by Elysium on Monday 15th March 20:51
Elysium said:
On average PCR is finding 25 times more cases per test than LFD.
This could be about the difference between symptomatic and asymptomatic prevalence.
Or it could mean PCR is overly sensitive
Or it could mean that LFD is no sensitive enough
Or it could mean that infectious asymptomatic cases are rare.
In any event average LFD positivity for the whole of March, including the school testing is just 0.09%
9 cases per 10,000 tests.
Mind blown.
I doubt PCR is 'overly' sensitive - in a testing sense, Covid is digital - you've either got it or you haven't (like an elephant). This could be about the difference between symptomatic and asymptomatic prevalence.
Or it could mean PCR is overly sensitive
Or it could mean that LFD is no sensitive enough
Or it could mean that infectious asymptomatic cases are rare.
In any event average LFD positivity for the whole of March, including the school testing is just 0.09%
9 cases per 10,000 tests.
Mind blown.
- Whether you've enough to then pass it on is a second question
- Whether you were tested at the right time in the development of Covid in your system is a third.
LFD lack of sensitivity...I have heard that. I have also heard (don't shoot me if this is wrong) it's possible to mis-perform the test, i.e. get a false negative...much like the old do-at-home brain-ticklers which a lot of people didn't shove far enough up their nose.
Infectious AND asymptomatic...in a primarily aerosol infection-environment (i.e. non-contact with others, given all the social-distancing measures currently in place), from what I know that would be in line with other coronaviruses* - if you're not coughing or sneezing then unless you're licking your fingers and touching everything it's difficult to see an infection vector.
Final thought - is there any difference in the 'selection methodology' - now we've a cheap and 'easily-performed' test is there a much higher proportion of random testing (asymptomatic) vs necessary testing, also is there a much higher proportion of testing outside of medical / care environments (so of people in lower-risk occupations)?
* Cold & flu, I mean.
havoc said:
I doubt PCR is 'overly' sensitive - in a testing sense, Covid is digital - you've either got it or you haven't (like an elephant).
- Whether you've enough to then pass it on is a second question
- Whether you were tested at the right time in the development of Covid in your system is a third.
LFD lack of sensitivity...I have heard that. I have also heard (don't shoot me if this is wrong) it's possible to mis-perform the test, i.e. get a false negative...much like the old do-at-home brain-ticklers which a lot of people didn't shove far enough up their nose.
Infectious AND asymptomatic...in a primarily aerosol infection-environment (i.e. non-contact with others, given all the social-distancing measures currently in place), from what I know that would be in line with other coronaviruses* - if you're not coughing or sneezing then unless you're licking your fingers and touching everything it's difficult to see an infection vector.
Final thought - is there any difference in the 'selection methodology' - now we've a cheap and 'easily-performed' test is there a much higher proportion of random testing (asymptomatic) vs necessary testing, also is there a much higher proportion of testing outside of medical / care environments (so of people in lower-risk occupations)?
* Cold & flu, I mean.
The interesting thing is that the positivity rate of PCR and LFD tests has moved together. Which suggests that what we are seeing is not particularly due to changes in the 'blend' of asymptomatic or symptomatic people attending for testing. - Whether you've enough to then pass it on is a second question
- Whether you were tested at the right time in the development of Covid in your system is a third.
LFD lack of sensitivity...I have heard that. I have also heard (don't shoot me if this is wrong) it's possible to mis-perform the test, i.e. get a false negative...much like the old do-at-home brain-ticklers which a lot of people didn't shove far enough up their nose.
Infectious AND asymptomatic...in a primarily aerosol infection-environment (i.e. non-contact with others, given all the social-distancing measures currently in place), from what I know that would be in line with other coronaviruses* - if you're not coughing or sneezing then unless you're licking your fingers and touching everything it's difficult to see an infection vector.
Final thought - is there any difference in the 'selection methodology' - now we've a cheap and 'easily-performed' test is there a much higher proportion of random testing (asymptomatic) vs necessary testing, also is there a much higher proportion of testing outside of medical / care environments (so of people in lower-risk occupations)?
* Cold & flu, I mean.
The positivity of both tests has gradually declined since New Year with PCR overall finding 25 times more cases than LFD.
To the extent that the total LFD cases to date is only 2/3rds of what we were finding in a single day through PCR.
Excess deaths update with todays ONS figures for the Wk ending 5th March:
1. The big news is that we no longer have statistically significant excess death. 11,592 deaths in total for week 9 compared to a 5 year average of 11,322. Below the 2015-19 maximum of 11,780:
So the number of people dying is essentially 'normal':
2. ONS reported 1,685 deaths 'due to' COVID in wk 9. So this is only 'normal' because deaths from other causes are at an historic low. I believe this is now concusive evidence of the re-badging of deaths due to our methodology.
3. This graph shows the same comparison with 'statistically significant' excess deaths. I would argue this is the clearest way to compare the first and second waves. The scale of the second wave is now increasingly disconnected from unusual levels of death:
1. The big news is that we no longer have statistically significant excess death. 11,592 deaths in total for week 9 compared to a 5 year average of 11,322. Below the 2015-19 maximum of 11,780:
So the number of people dying is essentially 'normal':
2. ONS reported 1,685 deaths 'due to' COVID in wk 9. So this is only 'normal' because deaths from other causes are at an historic low. I believe this is now concusive evidence of the re-badging of deaths due to our methodology.
3. This graph shows the same comparison with 'statistically significant' excess deaths. I would argue this is the clearest way to compare the first and second waves. The scale of the second wave is now increasingly disconnected from unusual levels of death:
Elysium said:
havoc said:
I doubt PCR is 'overly' sensitive - in a testing sense, Covid is digital - you've either got it or you haven't (like an elephant).
- Whether you've enough to then pass it on is a second question
- Whether you were tested at the right time in the development of Covid in your system is a third.
LFD lack of sensitivity...I have heard that. I have also heard (don't shoot me if this is wrong) it's possible to mis-perform the test, i.e. get a false negative...much like the old do-at-home brain-ticklers which a lot of people didn't shove far enough up their nose.
Infectious AND asymptomatic...in a primarily aerosol infection-environment (i.e. non-contact with others, given all the social-distancing measures currently in place), from what I know that would be in line with other coronaviruses* - if you're not coughing or sneezing then unless you're licking your fingers and touching everything it's difficult to see an infection vector.
Final thought - is there any difference in the 'selection methodology' - now we've a cheap and 'easily-performed' test is there a much higher proportion of random testing (asymptomatic) vs necessary testing, also is there a much higher proportion of testing outside of medical / care environments (so of people in lower-risk occupations)?
* Cold & flu, I mean.
The interesting thing is that the positivity rate of PCR and LFD tests has moved together. Which suggests that what we are seeing is not particularly due to changes in the 'blend' of asymptomatic or symptomatic people attending for testing. - Whether you've enough to then pass it on is a second question
- Whether you were tested at the right time in the development of Covid in your system is a third.
LFD lack of sensitivity...I have heard that. I have also heard (don't shoot me if this is wrong) it's possible to mis-perform the test, i.e. get a false negative...much like the old do-at-home brain-ticklers which a lot of people didn't shove far enough up their nose.
Infectious AND asymptomatic...in a primarily aerosol infection-environment (i.e. non-contact with others, given all the social-distancing measures currently in place), from what I know that would be in line with other coronaviruses* - if you're not coughing or sneezing then unless you're licking your fingers and touching everything it's difficult to see an infection vector.
Final thought - is there any difference in the 'selection methodology' - now we've a cheap and 'easily-performed' test is there a much higher proportion of random testing (asymptomatic) vs necessary testing, also is there a much higher proportion of testing outside of medical / care environments (so of people in lower-risk occupations)?
* Cold & flu, I mean.
The positivity of both tests has gradually declined since New Year with PCR overall finding 25 times more cases than LFD.
To the extent that the total LFD cases to date is only 2/3rds of what we were finding in a single day through PCR.
Fairly new study, although, very small!
Within the next few weeks everyone over 50 and those with vulnerabilities of all ages will have been vaccinated. The latest view is that this might reduce their risk of serious illness or death by 70%+.
This graph shows ONS weekly COVID-19 deaths taken straight from their spreadsheet. The data wrapper link takes you to an interactive version, which you will also reach if you click on the link:
https://datawrapper.dwcdn.net/7N0Lb/1/
It's necessary to compress under 40's into a single age band to make it visible on this scale.
Now imagine all age bands over 50 reducing by 70%
This graph shows ONS weekly COVID-19 deaths taken straight from their spreadsheet. The data wrapper link takes you to an interactive version, which you will also reach if you click on the link:
https://datawrapper.dwcdn.net/7N0Lb/1/
It's necessary to compress under 40's into a single age band to make it visible on this scale.
Now imagine all age bands over 50 reducing by 70%
Gassing Station | News, Politics & Economics | Top of Page | What's New | My Stuff