When testing the assumption for linearity with the log odds in my binary logit model (using Box-Tidwell), I found that one of my predictors does not satisfy this assumption. What should I do next? Transform this predictor? In what way? And how to do that?
PS: I am using RStudio.
data-transformation. Your understanding of the subject matter should be the first source of suggestions about ways to transform predictor variables. – EdM May 03 '18 at 21:54