Is there a correct method to center (mean or median centering) a skewed predictor variable in a multilevel model? The predictor is a skewed, count variable and will feature in a two-way interaction with a binary predictor (primary interest for the study). In most multilevel models, centering continuous predictors is commonplace when dealing with interactions, but I'm not sure if this is appropriate with a highly skewed positive predictor.
I have found the following links on crossvalidated, but they don't quite address my question completely:
https://stats.stackexchange.com/questions/216195/centering-skewed-predictors which suggests that median centering is also appropriate.
