For anyone who wants to get started, I’ve opened a PR on posteriordb that contains a bunch of models that have already been tested thoroughly, as they were used to evaluate whether covariance functions developed in C++ were good. It contains a lot of the demo models when the Stan group started transferring and trying to integrate from GPstuff:
But I believe in order to get this merged, a few things need to happen:
- syntax: I think the names of some of covariance functions have changed and there’s been other changes to the syntax
- you need to use R to produce the proper stan data file that get uploaded to the repo so that the models can be run automatically
- you’re going to have to split some up, for example the survival models I used different likelihood functions in the same file, to make it easier to develop.
- you get the point.
The models are valid, I can just share a google drive that has all the R and stan code, and then since rstan has been updated since then, you’ll need to update the R code so that cmdstanr or rstand commands work, and then upload the data or simulated data.
You’ll learn git, basic R, basic Stan, running models and diagnostics, and basics of software dev.
If anyone’s interested, I’ll share a google drive with all of the components. You have to dig through it a bit, but I’ll walk you through it. This should get you more fluent at how to contribute to a github repo.
(Honestly, something I just don’t feel like doing). Just ping me if you’re interested, I can walk you through it.