Hello Statalist,
I have a fairly complicated Cox regression model in that the dataset is
augmented to account for competing mortality risks (using the data
duplication method of Lunn and McNeil 1995) and, further, the model is
stratified by 4 different populations. Therneau and Grambsch (2000) say
that is inappropriate to use the averaged schoenfeld residuals for
stratified models, in particular when there are strata x covariate
interactions (which I have). They say this is a limitation as implemented
in SAS and SPLUS because both programs provide an overall estimate of
variance averaged across risk sets. Do you know if Stata provides the same
when strata are specified? They recommend a workaround that I can't seem to
replicate in Stata so perhaps you can help me there as well. According to
Therneau and Grambsch (2000) all one needs to do is:
1) fit the overall model to the data
2) refit each stratum separately using (iter=0 and initial options in SAS or
SPLUS) to force the same coefficient estimates as the overall fit.
3) Extract and save the scaled schoenfeld residuals
4) combine saved results into a single matrix and proceed with the usual
tests and plots.
While the process should be simple, I am hung up on steps 2 and 4. Might
anyone with a bit of knack for working beyond the canned command lines be
able to help me with a bit of cookbook to work through this process?
Thanks much,
jacqui
********************************************
Jacqueline L. Frair
Ph.D. Candidate
Department of Biological Sciences
University of Alberta
Edmonton, Alberta T6G 2E9
PHONE: 780-492-0083
FAX: 780-492-9234
EMAIL: [email protected]
********************************************
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
