Selecting predictors
•
Selection depends strongly on dataset
–
Sensitive to individual patients in set
Sensitive to population statistics
Robustness of model selection: bootstrapping