Hi,

I'm relatively new to Stata and am trying to set up some data for a Cox regression. I have about 500 subjects who have data recorded for either right or left, or both eyes, which will obviously form correlated observations. The 'event' is a pressure rise inside the eye and we are looking for that in any eye that had data recorded (i.e. the fellow eye was not used specifically for some form of matching).

I'm not sure I understand the difference between using stset with the id option, or just including the vce(cluster id) option following stcox to account for the clustering in the analysis (and not using stset id).

When specifying id in stset, I receive the following warnings, so the next step is to work out what they mean if I should do this.

Thank you,

Paul

HTML Code:
686  total observations
         12  multiple records at same instant                   PROBABLE ERROR
             (NdaysRX[_n-1]==NdaysRX)
          2  observations end on or before enter()
         32  observations begin on or after (first) failure