Yep, these are raw counts of deaths and cases, the idea being to look for homologous patterns of variation between death counts and case counts. It's the pattern that matters rather than absolute or scaled numbers. Corrections for double counting of cases will not be possible unless they release such data. There is a major headache with defining what constitutes double-counting in that folk will get tested under pillar 1 or pillar 2 or both, and they may or may not have multiple test results under one or more schemes. Then there's deciding the exclusion window to avoid the same person presenting twice - should this be a week, fortnight, month or something else? However this doesn't really matter providing double counting of cases has been a constant factor over time.
That 1st wave really does look like the curve predicted using a simple SIR model or a Gompertz curve.
Whilst writing, thank you so much for your explanation and custom analysis of ARIMA (re Hadcrut).
My pleasure. Yep, it is as Gompertz as they come. I have a little study lined up along these lines but just need to find the time.
Is this raw counts? What if it was cases per 100k ? As corrected for double counting etc.
Yep, these are raw counts of deaths and cases, the idea being to look for homologous patterns of variation between death counts and case counts. It's the pattern that matters rather than absolute or scaled numbers. Corrections for double counting of cases will not be possible unless they release such data. There is a major headache with defining what constitutes double-counting in that folk will get tested under pillar 1 or pillar 2 or both, and they may or may not have multiple test results under one or more schemes. Then there's deciding the exclusion window to avoid the same person presenting twice - should this be a week, fortnight, month or something else? However this doesn't really matter providing double counting of cases has been a constant factor over time.