Does normalizing your data affect outlier detection?