No issue just a comment to say that this is a lovely set-up repo and the model itself looks nice both in terms of implementation and method.

I will probably have a run-through of it in the next few days as TBH I think these kinds of regression approaches are the way to go as so much cleaner to implement than doing the decomposition into the forecast and delay dist I am doing.

Being able to define everything in terms of multi-dimensional splines for example is very nice (especially not having to try and get them working in stan!). Have you explored 3 dimensional versions (that include the date of report for example)?

Only substantive comment I had from the code was to wonder if you have explored the other bam settings? I have had good experiences using samfrac to get initial conditions (though maybe not available using the discrete approach) and nthreads to distribute across more cores.

Sam