It looks like you have added lots of 0.5%s
No I certainly did not!
It just depends on the actual or real case fatality ratio being constant or a
fixed property of the virus.
It looks like 0.15% in the general community population from my calculations which is OK re John P.A. Ioannidis pitching it at around 0.3%
You can’t use the PHE type cases data directly by integrating it
as you would have to correct it for cases over tests
and more importantly there is heavy sample selection bias as well.
Thus on the 28th April the sample selection bias PHE type data .
4730 cases / approximately 38,000? tests = 12%
ONS for the general population has case /test of 0.5% ish
You could say the ONS community data is bias as well
which it is but the ‘population’ that it is derived from is much larger
eg everybody outside a hospital and care homes etc
Using ONS data will more likely to under estimate the % who have had it
I will go back to it and try and think about presenting it in another simpler way.
40% is ok the best set of data is the Charles de gaulle data were it stopped at 47%
and the T-cell immunity covering the rest.
that was on a packed ship with shared air conditions and food preparation etc
Ok I will do it bits.
We want a true fatality/cases value for the 26th April.
For that we need the true case number which will be approximately
The ONS case number + “Nosocomial” case number.
We can make wild guess’s at that Nosocomial” case number.
However,
[ It won’t affect % of the population who have had covid calculation much
except perhaps changing it to 37-45% maybe?- I will do that later ]
We assume that the average case fatality rate over the whole “epidemic” is close to the estimated value of the 26th . [why? I will do that later]
We use average case fatality rate over the whole “epidemic” to calculate number of people who would have been PCR case positive.
Is that ok so far ?
It just depends on the actual or real case fatality ratio being constant or a fixed property of the virus.
Fatalities are the real measure of covid-19, but we are investigating the total/false/real test positives. We are now moving on to see if we might have a stab at those who were infected and survived (a stab might finish them off properly).
It looks like 0.15% in the general community population from my calculations which is OK re John P.A. Ioannidis pitching it at around 0.3%
That's reasonable. A 0.15% prevalence will produce the current 2.5% total positives at a specificity of 97.62%, so this is quite possible (with a resulting 95% false positive rate, good grief! 0.3% would require 97.73% and 90% respectively.)
You can’t use the PHE type cases data directly by integrating it as you would have to correct it for cases over tests, and more importantly there is heavy sample selection bias as well.
Yes. Initially, testing was limited to those showing symptoms - hence up to 38% on 1st April. As testing has expanded, the selection is less discriminating and approaches a better representation of the whole population.
Thus on the 28th April the sample selection bias PHE type data .
4730 cases / approximately 38,000? tests = 12%
ONS for the general population has case /test of 0.5% ish
I agree on the 12%. What is the 0.5% ish? The earlist ONS estimate I find is 0.27% for 27 April to 10 May:
https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/coronaviruscovid19infectionsurveypilot/england14may2020
(This requires 93% specificity to match the 7% total positives at the time - and a 97% false positive rate!)
40% is ok the best set of data is the Charles de gaulle data were it stopped at 47% and the T-cell immunity covering the rest.
that was on a packed ship with shared air conditions and food preparation etc
I am not familiar with that, but I am curious as to how you are going to estimate the total who have been infected. It might, indeed, involve a bit of adding up the percentages as they are not normally re-tested (I may be wrong about that). If you could reasonably show that it may have passed through such a large sector of the population, it would be quite significant. So good luck with it!
OK I leaving the subject of false positives for the moment and planning to go back to return to it after resolving the case fatality rate.
I don’t think the ONS lab results have a false positive problem ; I think it is mainly with the pillar 2 PHE data.
ONS labs got down to 0.05% constantly for 6 weeks.
If that isn’t an empirical good false positive rate I do not know what is.
I think ONS labs may have been repeat testing positives to reduce false positives which is the obvious thing to do to eliminate a lot of them.
I do not believe pillar 2 labs have a ever had a false positive rate of 0.05%
I believe I can demonstrate what the pillar 2 false positive rate was from the 15th June to 30th July
“
What is the 0.5% ish? The earlist ONS estimate I find is 0.27% for 27 April to 10 May “
The spreadsheet for the ONS survey has gone from the internet they were supposed to be updating it today; it maybe offline because of that I have a saved a copy of the spreadsheet.
There is this that I have attempted to copy from it.
Figure 1: The most recent modelled estimate suggests the number of infections in England has increased in recent weeks
Estimated percentage of the population in England testing positive for the coronavirus (COVID-19) on nose and throat swabs since 26 April 2020
Modelling reported in the previous publication, provided to give longer term context
Date % testing positive for COVID-19 95% credible interval
Lower Upper
26 April 2020 0.45% 0.28% 0.68%
Date % testing positive for COVID-19 95% credible interval
Lower Upper
26 April 2020 0.45%
And the link ;
has still got some dinky graphs on almost the first page.
Modelled estimates are used to calculate the official reported estimate. The model works by smoothing the series to understand the trend and is revised each week to incorporate new test results.
I don’t want to get distracted with specifity sensitivity at this point.
……… I am not familiar with that, but I am curious as to how….
https://en.wikipedia.org/wiki/COVID-19_pandemic_on_Charles_de_Gaulle
It is a bit of a classic case study
I think what I said of it was basically correct, have used it before on another forum months ago.
…………. how you are going to estimate the total who have been infected. It might, indeed, involve a bit of adding up the percentages as they ….
No!
………….If you could reasonably show that it may have passed through such a large sector of the population, it would be quite significant. So good luck with it!................
Done it written it up , cross checked it , it still looks good.
If it was established in India that for every of 100 road traffic accident cases there was one road traffic accident death or 1% .
And in three months there was 37,000 road traffic accident deaths.
How many road traffic accident cases would there be?
OK, i'm just a builder from Walthamstow, albeit with a hard earned o level in English language (i'm quite the exponant) and a couple of cse in woodwork and metalwork, and all this talk is way above my pay grade. But i have to ask. Why ain't you lot advising the morons at Westminster?
