[Q] Customizable GAM - missing functionalities that exist in mgcv

[maayanamid]

In R, it is possible to create a distribution family and define multiple custom functions: https://stat.ethz.ch/R-manual/R-devel/library/stats/html/family.html I want to implement a conditional poisson distribution using pygam, but from what I understand from the docs I can only chose one of the given distributions, a custom log function and the given constraints. Is there a way to reach the full functionality of a custom distribution function, which would include: linkfun, d2link, d3link, d4link, linkinv, variance, dvar, d2var, d3var, dev.resids, aic, mu.eta? Thanks!

[dswah]

Hi @maayanamid no unfortunately not right now. perhaps in the future when we implement REML, then we will be able to accomodate this level of modularity. :)