Missing data and prediction modelling

Dear prof
with 20 % missing data in some candidate predictors, is it not better we fit the pre specified models in imputed datasets and pool the results or fit in the original data set and then do sensitivity analysis with imputed data(suggested by a methodologist here)? if using the imputed datasets, how to assess the calibration and discrimination etc?
thanks in advance

Please post this as a reply to the last post on RMS Missing Data - modeling strategy - Datamethods Discussion Forum and I’ll delete it here.

1 Like