OT - I read through that report and most of the info he has is accurate and useful.
But his projections and transmission rates are way off. You can validate that easily by checking his projections against the tracking sites.
My model basically says that every 8 days we have a 10x increase in total number of *reported* infections. At this rate, by Easter Sunday we will be reporting over a million infections in the US. If we fall short of that number, then you know that something has gone right (or that we have stopped testing and reporting). But so far there is no evidence that the rate is slowing down.
I expect that by this Friday we report a 100k. I suspect (and hope) that when that happens, everyone - starting with the White House - gets serious about mitigation and flattening the curve. Otherwise, the following weekend there will be a million infections.
If you look at Italy, they had a weekly rate of 10.8x (even worse than the US), but now they are at 2.4x weekly rate. This is still high. You have to bring it down to less than 1.4 per week and then 1. But 2.4 is hell of a lot better than 10.8.
The US also has the problem of not knowing how many are infected in the first place, so it is hard to tell how effective the current protocols are. So the only way to estimate that is to look at the mortality number. This is the other part of that research that I disagree with. I think the mortality lags by about a week - not more than 2 weeks as the report states. Again, my model matches the data, his doesn't.
Anyway, based on the mortality data, *for now* you can estimate the current number of infections as 175 x total mortality. In about 10 days to 2 weeks, this approximation will no longer work. My model is meant to be accurate enough during the early stages and then be refined as the parameters change. But for now, we can assume that the total number of infections in the US is ~83k - 110k of which only 44k have been discovered. This estimate matches early academic research that concluded the number of infections is at least 3x the reported number. |
