Preprocessing ahead of Predictor Screening analysis

Question

Hello,

I have some characterisation data that I would like to subject to predictor screening.  The variables are on different scales.  Is any preprocessing like standardisation required prior to the predictor screening?

Victor_G · Answer

Hi @kjwx109,

As @Byron_JMP mentioned, the Predictor Screening platform is based on a Random Forest, which is a tree-based method robust to outliers, collinearity between variables, scales of the variables...
You can find more infos here : Predictor Screening

If you need further infos or if you have more questions, don't hesitate :)

Byron_JMP · Answer

Tree methods, like predictor screening, partition, bootstrap forest, and boosted trees (and XGBoost) don't need pre processing.  They work well on dirty data too (missing, outliers, strong X correlation with Y...