Worth a read. The apparent (unless James is missing something obvious and I don't think he is) lack of calibration in some of the epidemiological modelling is very strange. I can't understand it & would love an expert explanation as to why this is the case https://twitter.com/jamesannan/status/1250143096735752193?s=19

Its particularly odd as it's trivial to do and seems like a fairly obvious activity to carry out. This makes me wonder whether there is some amazingly well founded theoretical reason why not and we non experts have totally missed this.

I generally like to let experts do their expert thing, so this is a bit uncomfortable for me. But James's simple model is, well, better (at forecasting mortality) than the Imperial modelling that has fed into the Govt response. It's a bit disconcerting to say the least.

Maybe this is like some of the CMIP6 models where everyone insists they are better because they simulate some stuff nicely even though they do a worse job on the historical temperature record, which is the main thing people care about. LOL.

(sorry)

Just kidding climate modellers!