11-14-2011 05:28 PM
I am using PROC PHREG to analyze survival times (time until death) of planted trees. I have run into several issues including extremely high hazard ratios and standard errors, and questionable parameter estimates.
I have two fixed categorical treatments (seven and four levels respectively) and I am anticipating significant interaction between the two treatments. I believe my problems are the result of the following two issues.
1. I have combinations of treatments levels where 100% of the planted trees survived (no event occured for a subset of trees). For these combinations of treatments 100% of the data is right censored.
2. I have a fractional factorial design. Not all levels of my second treatment were set up for all levels of my first treatment.
Is it possible to analyze my data in their current form using PHREG or any other PROC. I have successfully analyzed a subset of the data when I eliminated the above problems. Any help is much appreciated.