A tale of two analyses:
-
A tale of two analyses:
I've mentioned several times that I believe the PMC19 estimates for #COVID19, which many take as gospel, are inflated. Yesterday, PMC (Michael Hoerger) estimated 200,000 new COVID infections daily in the US. At the same time, JPWeiland estimated 90,000 daily infections. PMC thinks 1 in 244 is actively infectious, while Weiland estimates 1 in 750.
Why such a disparity, and who's correct?
https://x.com/JPWeiland/status/2047800248643223657
1/3
-
A tale of two analyses:
I've mentioned several times that I believe the PMC19 estimates for #COVID19, which many take as gospel, are inflated. Yesterday, PMC (Michael Hoerger) estimated 200,000 new COVID infections daily in the US. At the same time, JPWeiland estimated 90,000 daily infections. PMC thinks 1 in 244 is actively infectious, while Weiland estimates 1 in 750.
Why such a disparity, and who's correct?
https://x.com/JPWeiland/status/2047800248643223657
1/3
Both PMC and Weiland use the same data source—wastewater data. Yet both arrive at different estimates.
We don't know which is correct (because, of course, there's no accurate data on actual infections). But, as someone who's followed COVID data closely for years, I trust Weiland's estimate much more.
For evidence, look at PMC's forecast published just two weeks ago (page 12: https://pmc19.com/data/PMC_COVID_Report_April132026.pdf.) Its forecast was for level infections and 250,000-300,000 new daily infections today.
2/3
-
Both PMC and Weiland use the same data source—wastewater data. Yet both arrive at different estimates.
We don't know which is correct (because, of course, there's no accurate data on actual infections). But, as someone who's followed COVID data closely for years, I trust Weiland's estimate much more.
For evidence, look at PMC's forecast published just two weeks ago (page 12: https://pmc19.com/data/PMC_COVID_Report_April132026.pdf.) Its forecast was for level infections and 250,000-300,000 new daily infections today.
2/3
Instead, we have declining risks and, by PMC's current estimates, 20-33% fewer infections than forecast.
This isn't an anomaly for PMC. If you go back and look at the past forecasts, PMC systematically overestimates future risks. That calls into question all the estimates on the site (in particular, its estimate that the average American has had 5.17 infections).
I believe PMC's consistent history of overestimating forecasts demonstrates that its model has a pervasive bias.
3/3
-
R relay@relay.mycrowd.ca shared this topic
-
Instead, we have declining risks and, by PMC's current estimates, 20-33% fewer infections than forecast.
This isn't an anomaly for PMC. If you go back and look at the past forecasts, PMC systematically overestimates future risks. That calls into question all the estimates on the site (in particular, its estimate that the average American has had 5.17 infections).
I believe PMC's consistent history of overestimating forecasts demonstrates that its model has a pervasive bias.
3/3
@augieray
Augie,I understand your analysis but am confused by your thread.
I didn't understand why you conclude one is more accurate than the other. The previous forecast is in the ballpark for the current number so the model is somewhat consistent within the probably huge error bars.
My operating perspective is the two models give an upper and lower model bound, and the likelihood is the answer lies in between. But we cannot know for sure either way - they both could be wrong.
I also watch Tara Moriarty's work which was calibrated to excess deaths.
I watch the trends more than the absolute risk numbers as a result, and use the models to bound the absolute risk, hoping they are in the ballpark.
Could you elaborate on the basis you use to prefer one over the other?
-
@augieray
Augie,I understand your analysis but am confused by your thread.
I didn't understand why you conclude one is more accurate than the other. The previous forecast is in the ballpark for the current number so the model is somewhat consistent within the probably huge error bars.
My operating perspective is the two models give an upper and lower model bound, and the likelihood is the answer lies in between. But we cannot know for sure either way - they both could be wrong.
I also watch Tara Moriarty's work which was calibrated to excess deaths.
I watch the trends more than the absolute risk numbers as a result, and use the models to bound the absolute risk, hoping they are in the ballpark.
Could you elaborate on the basis you use to prefer one over the other?
@EricCarroll Sorry for being confusing. Tough to provide thorough analysis in 500-character posts.
For years, PMC has consistently overestimated future risks. Its forecasts are ALMOST always higher than what we experience. That suggests Hoerger's model consistently inflates infections based on wastewater data.
In the end, you're right—we can't know. But, tracking positive rate and hospitalizations, Weiland's forecasts and analysis always seem on point, not just a "lower model bound."
-
Instead, we have declining risks and, by PMC's current estimates, 20-33% fewer infections than forecast.
This isn't an anomaly for PMC. If you go back and look at the past forecasts, PMC systematically overestimates future risks. That calls into question all the estimates on the site (in particular, its estimate that the average American has had 5.17 infections).
I believe PMC's consistent history of overestimating forecasts demonstrates that its model has a pervasive bias.
3/3
5.17 sounds accurate for my kids, across 2 blended households. Which is terrifying for our future.
Adults in each affected household seem to have fared slightly better.Which feels "weird" but would align well with the statistic of ~50% transmission within a household.
-
R relay@relay.infosec.exchange shared this topic
