hi, I have been working on the generalized additive model from a machine learning perspective for a month or so. I have understood the high-level working of such models and so on so. But I don't understand how the p-value is determined in a penalized version of GAM model, as that becomes a non-parametric version. Although I have a fair idea on determining p-values in parametric models, I don't understand the non-parametric settings.

What I plan to discuss here is what are the theories associated with p-value calculation in such cases. The goal of my understanding is to correct a known bug https://github.com/dswah/pyGAM/issues/163 in the python package for gam modeling.

thanks in advance for your valuable contributions to the discussion.