Thanks, Samuele. This is brilliant. Clearly the cut off assumption plays a big role here in generating the predictor matrix. Do you have any recommendations in this regard - perhaps trying a bunch of cut offs, averaging, etc and the impact on the outcome, if any?