Displaying 1 result from an estimated 1 matches for "reformat_dataset".
2013 Jul 06
1
problem with BootCV for coxph in pec after feature selection with glmnet (lasso)
...ze.
However, once pec refits the model its size changes (356 -> 318 variables).
Apparently something is happening during the bootstrap sampling that removes
some variables.
As mentioned, I used na.omit in the preprocessing so there should not be any
NAs.
Here are some details from my workspace:
Reformat_dataSet: 368 obs. Of 356 variables
print (glmnet.cox) ----> 354 df, p=1 n= 368, number of events= 288 (354 df
= 354 variables + time and status => 356 variables)
Here is the pec function and the error:
pec.f <- as.formula(Hist(time,status) ~ 1)
brierGlmnet <- pec(list(glmnet.cox), data = refo...