Excuse my dumb question, but I did an univariate logistic regression where the sign of the coefficient of my variable was negative (and it was significant). Once I have input it into a multivariate regression, the sign of the coefficient of the same variable became positive (still significant). Why does this occur? Is it because of outliers or collinearity between variables (even though as I said the variables are significant), or is it something more subtle?
Asked
Active
Viewed 695 times
