On Dec 22, 12:00 pm, citam.s...@GMAIL.COM (Kevin Viel) wrote: > Greetings, > > Might anyone be using the counting process style with time-varying > covariates? If so, I would appreciate any comments or references for the > martingale residuals (RESMART in the OUTPUT statement). In particular, > what is the implication of having a residuals for multiple intervals for a > given subject? > > Thank you in advance, > > Kevin

Never have - won't be much help. Looks like the counting process input is designed for recurrent events that may or may not be associated with "time-varying" covariates - not be confused with "time- dependent" covariates that violate the PH assumption while remaining constant through the study. Seems like an intuitive way to deal with time-varying covariates too, even without recurrent events, yes? Maybe this is what you're doing. Seeing Therneau as the designerof the input matrix and looking at the Therneau and Garambsch text for extending Cox models, I'm thinking this fellow knows a bit about your problem. His text seems pretty thorough, but expensive, so you might start with his website: www.mayo.edu/hsr/biostat.html - I don't know but the website may just be good for macros and datasets associated with the text. The text seems to have much more on martingale residuals than Schoenfelds. good luck. Shawn