Smooth functions of a variable - modeling syntax

[wesleyburr]

Smooth functions of a variable - modeling syntax

[rmcelreath]

Yeah good point. It's an awkward problem either way, given that typed variables are not so natural to so many students. I could give your idea about %*% a try. Should we easy to add a check for uncormforbale objects while the linear model is parsed.

[wesleyburr]

Update on this: I asked the students to try to use ulam() this week, and ran into a similar issue. It seems like the way you parse the formulas into Stan code is a bit different for ulam() [seems = it is, as you clearly state in the docs, because of the more powerful support for GLMMs and other features].

In an ulam() call, setting the start values is insufficient to properly set up the vector size for the parameters for a smooth function. The workaround I found is to manually drop into Stan-like syntax and pre-pend the prior statements for those particular (vector) priors, e.g., vector[9]:w ~ rnorm(0, 1) instead of the more textbook-described w ~ rnorm(0, 1), where the parser takes care of it.

I have some students who seem to like to push the boundaries ....