time step documentation?

[bbolker]

I'm currently trying to port an example over from the fitode package.

In fitode, the times and time steps are automatically taken from a column in the data set. In this case, the time units are in years. The rate parameters c(beta=70, gamma=60) are sensible in terms of years, but not in terms of time steps (which are actually weekly). Obviously I can just multiply by 1/52, but I'm wondering if there is now a "set delta-t" functionality (there is a promising sounding mp_time_scale() function, but it's undocumented and the code is not transparent -- I can guess the existence of internal functions Daily, Weekly, etc., but ???

While we're at it, are there examples yet of lining up calibrations with (say) multi-stream data that have different missingness patterns ... ? (update: I see there is something here, but it looks incomplete — comment says "## won't work yet ... need more sophisticated obs/sim matching ...")

I see there is a delta-t pull request but not sure about status (the last GH actions check was cancelled ...)

[stevencarlislewalker]

As usual I'm having trouble keeping up with documentation, internal and otherwise.

Time Steps

Here's a link to the success that I had on the #314 pull request: #313 (comment) . I'm feeling a little wimpy about merging into main, but I'll try again today.

This test compares RK4 on #314 with deSolve.

The mp_time_scale() function should be marked experimental, which I'll do today. Daily and Weekly are similarly a bit of a mess.

Multi-stream Data with Different Missingness Patterns

Using mp_tmp_calibrate() makes this easy (for me). The techniques in Advanced Calibration continue to get outdated and I'm having trouble keeping up. I'll improve the docs a bit today.