I don't understand your table or summary results. It has the same n as "symptomatic" (from part 4) which came in at 28% of all COVID but this subset of "acute symptomatic" comes in at 45%?
OK, but still, how does the subset represent a larger portion of the overall population in your dataset? If "symptomatic" COVID is 28%, surely, "acute symptomatic" COVID must be a smaller percentage?
As Joel may be mentioning, seems like there is some error here.
Proportions seem like a hopeless quest due to Trust not having testing rates. Seems like only country-wide evolution of IFR is doable - either through just lifting it from a paper, or constructing it some place where testing rates are known.
Since the common denominator here is folk who've been tested at least once and found to be positive then it's not so hopeless unless you argue that clinically symptomatic cases were subject to a battery of tests until they found a positive result, or that asymptomatic but suspected cases were tested ad infinitum compared to symptomatic until they obtained a positive result. All this rests on the assumption that a positive test result alone is driving the ICD10 coding but this is not necessarily the case. We also have to explain why test bias has changed over time in this manner, if it indeed has. If bias has remained just as problematic over time then hopelessness melts into a warm puddle.
p.s. there's no error - Joel didn't realise 'n' pertains to the number of weeks in the sample, a data record being a week and not a case. An additional column of summed case counts is needed - I may well revise the tables with these when I can find time.
Shouldn't proportion of acute asymptomatic covid deaths be lower overall than proportion of any symptomatic covid deaths by definition? You report the opposite.
I don't understand your table or summary results. It has the same n as "symptomatic" (from part 4) which came in at 28% of all COVID but this subset of "acute symptomatic" comes in at 45%?
n = weeks not people! The data record is the week and not the case. To get to n = people we have to sum counts over weeks.
OK, but still, how does the subset represent a larger portion of the overall population in your dataset? If "symptomatic" COVID is 28%, surely, "acute symptomatic" COVID must be a smaller percentage?
An error in my batch code - part 4 is now revised with the corrected figures!
Damn! Had to revise my article that referenced it toohttps://metatron.substack.com/p/covid-the-biggest-threat-faced-in then!
This is what happens if I don't stuff carbs and reduce my bacon ration. Your 50% guesstimate was right on the money!
I've stared at enough mortality series to do this stuff in my head!
I've added an extra slide at the end of part 5 so we can see the two series bouncing along, this being penance for my shoddy coding!
As Joel may be mentioning, seems like there is some error here.
Proportions seem like a hopeless quest due to Trust not having testing rates. Seems like only country-wide evolution of IFR is doable - either through just lifting it from a paper, or constructing it some place where testing rates are known.
Since the common denominator here is folk who've been tested at least once and found to be positive then it's not so hopeless unless you argue that clinically symptomatic cases were subject to a battery of tests until they found a positive result, or that asymptomatic but suspected cases were tested ad infinitum compared to symptomatic until they obtained a positive result. All this rests on the assumption that a positive test result alone is driving the ICD10 coding but this is not necessarily the case. We also have to explain why test bias has changed over time in this manner, if it indeed has. If bias has remained just as problematic over time then hopelessness melts into a warm puddle.
p.s. there's no error - Joel didn't realise 'n' pertains to the number of weeks in the sample, a data record being a week and not a case. An additional column of summed case counts is needed - I may well revise the tables with these when I can find time.
Shouldn't proportion of acute asymptomatic covid deaths be lower overall than proportion of any symptomatic covid deaths by definition? You report the opposite.
It should - just going to check the batch code...
Bottoms, bottoms, bottoms and bottoms - revised figures coming out shortly.
My batch code for part 4 had errors - a revised article is now ready for your eyeballs.
I've added an extra slide at the end of part 5 so we can see the two series bouncing along, this being penance for my shoddy coding!
I've added an extra slide at the end of part 5 so we can see the two series bouncing along, this being penance for my shoddy coding!