How to add support for a new model
saudiwin opened this issue · 4 comments
Hey Vincent, it's me, your favorite annoying American programmer!! 😅
As I mentioned on Twitter, I just came up with an MLE implementation of ordbetareg using scipy. Estimation of marginal effects with numerical differentiation should be straightforward as it's a standard MLE fit with Hessian, etc. However, there does not yet seem to a be a protocol for adding support for models to the pymarginaleffects package as there is with the R version. Any particular way you want me to go about doing this? I'm not an expert python programmer by any means, but chatGPT has proven to be remarkably helpful...
Cool cool. I think the best place to look for an example might be here:
https://github.com/vincentarelbundock/pymarginaleffects/blob/main/marginaleffects/model_pyfixest.py
That looks great.
Would you want to define that as a general model class? I.e. if class = MarginalEffectsModel and all the methods are implemented, then it could pass the model sanity check.
Practically should I do a PR for this? I think it has to be within the codebase as I can't get around the model check otherwise.
Yes, it would need to be a class like ModelOrdBetaReg, which inherits from ModelAbstract, as in the example I showed you.
A PR would be great, because otherwise I probably won't have time to implement this in the near future.