Then ignore them. I also posted actual data, so feel free to stick to that if you’d like to have this discussion.
Why do you find it hard to believe? While we don’t know, I certainly find it believable. Youyang Gu, who is the modeler I most trust, thinks that July 10th was the peak for the actively infected, but his narrow confidence intervals still allow for April to have been higher. Many areas didn’t lock down much until near the end of March and there was almost no testing being done. Over 36% of the US deaths and 28% of the reported cases prior to May 1st were in New York, the place where they were sending sick people into nursing homes. We had a positivity rate that was 4 times what it is now. Was a lot of that simply testing bias? Absolutely! Do we know what the numbers were? Not even close. We still suck at it, but our testing is quite a bit more comprehensive than it was six months ago and our positivity rate is a bit more stable. That is why I picked early June as a starting point.
You do you. I’ll stick to seeing what I can derive from data that is far from perfect, as that’s what we have for now.
FWIW this resource, Oxford’s Center for Evidence Based Medicine’s continually updated review about CFR and IFR, has been shared before on this forum before. The section “Estimating COVID-19 Infection Fatality Rates (IFR)” is most useful.
Bolding below mine.
From the start determining the true IFR for each cohort, by age, by comorbidities, by other potential confounders like ethnicity, SES, location be it urban vs rural or home vs. in a congregate living facility or country, has been the big bugaboo. Even At NO point has overall CFR been a reasonable measure for that understanding of true IFRs. Even if one accepts that overall there are ten, or twenty, or whatever multiple of actual infections to those diagnosed, one cannot accept that that multiple is consistent across time periods or locations.
Cross referencing with positivity rate can help get some handle, so the high positivity rates early on mean that the number of true infection was being particularly underrepresented by the number of diagnosed cases.
The groups getting infected the most where is likely the biggest driver of IFR, to the degree that CFR reflects it.