I'm trying to build a classifier using a logistic regression and statsmodel is telling me that there is an issue of quasi-separation.

Well this isn't an issue! This is exactly what i'm trying to do: predicting the Endogenous Variable using the Exogenous Variable!

The real problem is that i can't find the output telling me what variable cause it and what are the associated values. Can anyone hint me where to find that?

Here is my error message

Possibly complete quasi-separation: A fraction 0.62 of observations can be perfectly predicted. This might indicate that there is complete quasi-separation. In this case some parameters will not be identified.

I'm using Python 3.4 and statsmodel 0.6.1

  • $\begingroup$ See stats.stackexchange.com/a/137547/17230. And it's "quasi-complete separation". $\endgroup$ – Scortchi - Reinstate Monica Feb 25 '16 at 22:32
  • $\begingroup$ I uploaded my Error Message. I understand that in the post you linked that they are talking about collinearity between exogenous variables. But I wouldn't think i have the problem... $\endgroup$ – MastaJeet Feb 25 '16 at 22:40
  • $\begingroup$ Is this raising an exception or just print a warning? If you have the results, then I think there should be some parameters with large values or very large standard errors. If that's the case, then these parameters would not be identified because of quasi-complete separation. $\endgroup$ – Josef Feb 25 '16 at 23:12
  • $\begingroup$ I wasn't linking to the question, but to my answer, which references a linear programming algorithm to identify which predictors cause separation. $\endgroup$ – Scortchi - Reinstate Monica Feb 26 '16 at 9:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.