Well, but now that he knows that deaths will only be in the hundreds of thousands - that's not so many voters out of the 230 million eligible voters.
Trump has done several things that focus the deaths and hardships on States and voter populations that vote Democratic. He's going to kill fewer Republicans than non-Republicans. That appears to be non-random.
- - - -
Always remember: none of the detection numbers in the US are reliable. We do essentially no testing before clinical diagnosis based on self-presentation, and often not enough after.
That will be a bell curve rather than sigmoid and should be proportional to the rate of new infections in the population,
Can't be both.
Unless by "proportional" you mean something other than I assume - I assume a constant of proportionality.
OK, I presume the new cases curve will be a bell curve like the hospital admissions curve, won't it?
? I don't follow that assumption.
Even when the hospital admissions break from the "sigmoid" (I assume "logistic") curve to show a decrease rather than an asymptotic plateau, they won't plot as a bell curve (I assume Gaussian).
Assuming it's a perfect Gaussian curve, of course, which it isn't - but it's a reasonable approximation.
I'm not following this. Why do you guys think hospital admissions would approximate a Gaussian, on either side of the peak?
Note that hospital admissions are going to vary by circumstances other than infection and disease - including timing and accuracy of diagnosis, demographics of the sick, availability of a hospital bed, and so forth. In the US they will also vary by insurance coverage and the political allegiances of local government. Is that where the randomness assumption - the Gaussian - comes in?
Also: changing the parameters of these curves is the whole point of the pre-vaccination response - buying time for the medical care response, the vaccine development, the transition to different work organization, etc - but in the process invalidating the older projections. If we do damp the logistic, as we very much hope, we will produce a quasi-linear response wherever we are when we do that, even in places other than the former inflection point - including places earlier in the curve, as well as later.
So we have to be careful in interpreting - for example: we might be delaying, not decreasing, the eventual peak. It's important to get a handle on the new curve we have created via response measures, and predict from that rather than extrapolating from the old one.
