I am working with a dataset large enough to make necessary the use of alogorithms more advanced than stats glm in order to compute binomial regression models. glm takes over a week to compute.
Large fixed effects binomial regression in R suggests rxGlm and speedglm. I tried both and managed to achieve significant improvements in speed (down to under 2 hours). However, I am facing the following problem:
In both cases, I cannot use the effects function plot(effect()) to create interaction plots similar to the following ones, because effects cannot deal with these kind of objects and / or needed data is missing.
I already had posted a question with bounty on rxGlm (How to plot interaction effects from extremely large data sets (esp. from rxGlm output)), to no avail. So I guess there is no way to do so using rxGlm.
Now my question: can I plot such interaction effect terms using speedglm, and if so, how?