Decoding smooth fits in brms

The best I’ve got for the degrees of freedom issue is to use information criteria. The waic() function returns a p_waic estimate, and the loo() function returns a p_loo estimate. I believe both can be interpreted as the effective number of parameters, somewhat analogous to a degree of freedom.

Some of us have experimented with notation for models with smooth terms in this thread, but I’m not so sure that’s what you’re looking for.

2 Likes