I am discussing this on another thread, or was, but I realized it would get buried and is another topic. I have a series (31) likert scale predictors for logistic regression (so all variables are measured on the same scale). After a lot of consideration I decided to go with odds ratios to rate the relative impact of my predictors. But based on Box Tidwel 5 of my variables are non-linear (two are close in that they are just below .05 on the Box Tidwel test).
The question is how do you compare (can you compare) relative importance of linear and non-linear predictors on the DV. (By importance I mean had the most impact on the DV relative to other variables].
The question is how do you compare (can you compare) relative importance of linear and non-linear predictors on the DV. (By importance I mean had the most impact on the DV relative to other variables].