Thank you very much for your reply.
How is it possibe to approximate the max number of variables that should be
used considering the number of samples and censored data? Experimentally, I saw
that for my specific data, I have to reduce down to 120 variables. So I
wondered how it could be
It is quite possible that you are trying to find an MPLE where none
exists, in which case non-convergence is a good thing.
Using only 300 cases for 188 or 215 variables is asking a lot, especially
when 26% of them are only partially observed. Since you did not sign your
message we have no idea
hi,
I applied coxph to my matrix of 300 samples and 215 variables and got the
following error
Error in fitter(X, Y, strats, offset, init, control, weights = weights, :
NA/NaN/Inf in foreign function call (arg 6)
In addition: Warning message:
Ran out of iterations and did not converge