At the moment @kimbers I’d rather not publish them in public. The methodology is novel, and I update them daily for near-term (7-10 day) forward projection. But the numbers are terrible, and I think they destabiise. Should the epidemic look like it does not flatten, then I will do so, because at that point people will need some shock tactics. I started with the naive method in that paper two weeks ago just to get something out there. In the meatime I solved the much harder problem and I’ve been using my more accurate methodology for the past week. The forward 7-day prediction has an error (stdev log of 0.3). For an exponential problem, I’m very happy with that. I’m now solving longer-term projection, but the errors will blow up fast, so not too helpful.
7-day lower limits (below which we should be confident that intervention is happening)
UK 3700
US 16000
DE 1500
ES 19000
FR 4400
IT 21000
NL 1750
So if total deaths reported NEXT saturday are lower then these values, we have solid evidence that the curve is flattening. Finger’s crossed…