SIRD model with terrible chain mixing for starters

[breckbaldwin]

@jtimonen

@breckbaldwin

Is it silly to be trying to estimate the (I)nfected for day0?

If you really have no idea what the number of infected is at initial time point, and it could be anything between 0 and population size, then it sounds like the data is just very difficult to fit anything to begin with. It could help if I saw the data. It in general is not silly in my opinion to try and estimate the initial state and it works if the data is well behaved with respect to the model, but it might not be the case here.

The 2020 data are in the .stan source file in transformed data, graphed below, total population: 214,110,287