1

I am trying to understand complete and quasi complete separation in the context of logistic binary regression. However, I have not found a clear source. I know the seminal paper

https://www.jstor.org/stable/2336390

but the definition is a bit hard to understand as it corresponds to multinomial logistic regression.

Could someone point me to a reference for these definitions on binary regression or to put them here in an answer?

albert
  • 11
  • 2
    Binary logistic regression is equivalent to multinomial logistic regression with 2 classes/groups. So, the paper you posted already gives the answer. Fig. 1 shows an example in the binary case. Might also be worth noting that linear separability is a geometric concept, not defined in terms of any probabilistic model (e.g logistic regression). – user20160 Jul 26 '19 at 10:24

0 Answers0