In a logistic regression setting, can there be a relation between the fact that covariates are highly correlated and the problem of perfect(quasi-perfect) separation?

  • $\begingroup$ That's a good question, but I think one should try to formulate it in a more precise way. Correlation is a property of the covariates only whereas separation is a proporty of both covariates and response. So I'm pretty sure you can think of example with very correlated covariate and no separation at all and vice-versa. $\endgroup$
    – Pohoua
    Jun 21, 2020 at 12:22


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.