GLM.jl
GLM.jl copied to clipboard
Perfectly separable test
A WIP candidate for the solution discussed in https://github.com/JuliaStats/GLM.jl/issues/286
TODOs:
- Needs to automatically disabled for all but Bernoulli and Binomial models, or at least able to be overridden
- Extraction of
YandY_predinsideperfectlyseparable()needs to be handled properly - Review of method for efficiency
Closes #286.
I've added a test for this issue. There may be a better test for proving success with the perfectly separable example dataset.
Ok, those sound like good points @andreasnoack. I'm going to bow out and let someone with better understanding of the package implement the suggested method.
Stale