From mailing list
"I am creating Multinomial Logistic Regression Models using MNLogit from statsmodels.discrete.discrete_model. Occasionally, I receive NaN as the function's value after termination.
I have equivalent code in R, and I do not receive NaN values. Can anyone explain this behavior? Basically, I am trying to predict one of three categories from several explanatory variables that are continuous in nature."
I have the data for this but am unable to replicate as of now.
Closing this as a duplicate of #516. It comes from near perfect separation IIRC.