Response-effect model overparameterized?

[stouffer]

I'm pretty sure that the model that gets fit with cxr_er_fit() is overparameterized. See this paper and this paper for more details.

I can probably give some suggestions for how to modify the code to deal with this if you want. If nothing else, it should definitely help improve convergence during the optimization step.