This repo is a very, very brief sample of some code and analysis corresponding to the idLogit method described in these slides, as applied to Carroll & Verhaal's "Authentic Distilleries" binary choice data. This case study discussed briefly in the slides starting at slide 58.

The statistical method used is a "new" one, which I've called idLogit. It's not really new at all, in that idLogit is a variant of L1 or "LASSO" regularized logistic regression. It is new in that the specific regularization chosen is tailored to the nature of the data and the effects we hope to model. See also the idLogit package, my Stanford ICME presentation on idLogit and the associated notebook

I recommend simply reviewing the bootstrap.ipynb file for an overview. However this file simply analyzes data and results, and does not contain any of the analysis itself. If you want to see how the analysis was done, review idLogit.py and runner.sh.

NOTE: Running this code requires the underlying data and/or results. The underlying data I am not authorized to provide, and the results from statistical estimation aren't particularly suitable for inclusion in a repo. Reach out to me at morrowwr@gmail.com if you are interested.

I also include a short writeup (writeup.pdf) This provides a bit more context and description, especially about some of the plots in the notebook. It was, though, mainly intended to inform Carroll & Verhaal's paper drafting.