02-04-2016 10:43 AM
I have a question regarding hierarchical cox regression model.
When you have two variables for clustering, for example hospital and physician, and want to do time-to-event analysis,
is the following code appropriate? If not, is there any statement or procedure for this analysis?
PROC PHREG data=DATANAME covs(aggregate) covm;
model TIME.*EVENT(0)= Variables /
id hospital Physician;
02-04-2016 12:28 PM
The ID statement specifies variables that identify observations in plots and in output data sets, but the ID variable doesn't change the analysis.
From what little information you've provided, I suggest you look at the "Analysis of Clustered Data" example in the PHREG documentation, which uses a RANDOM statement. The doc for the RANDOM statement explains the statistical details of the model that is fit when you use the RANDOM statement.