Dear list, I am willing to perform a cox PH regression analysis on my "time to reach a severity score of 5". I have 600 observations and 50 predictor variables. So I first want to perform a log-rank test (using the survdiff() function) on each individual variable in order to select a subset of them for cox analysis. My question is: is it possible to control and obtain the False discovery rate of the log-rank test for each of the 50 predictor variables? Thanks in advance, David