I am estimating a competing hazards model by including a dummy variable for "protected" status and then interacting my this protected indicator with each covariate in the model (a la Lunn and McNeil Method A). A linear summary of my model would look something like this: y = I(protected=1)+B1x+B2(x*I(protected=1). I would like to use -stcurve- to graph these two categories, but the graph (and resulting baseline hazard estimates) seem not quite right when I enter the following:
stcurve, hazard at1(protected=1) at2(protected=0)
After much puzzling I believe that the the problem is that stcurve is not also estimating separate values for the interaction terms, instead setting them to the mean values overall. So I tried the following, where 1 category has a 1 for protected and positive values for the interaction terms, and the other has a zero for both the indicator and the interaction terms:
stcurve, hazard at1(protected=1 pctrawcount_prot>0) at2(protected=0 pctrawcount_prot=0)
Unfortunately Stata doesn't allow anything but "=" in the -at- statements.
Does anybody have a suggestion for how to work around this?
Many thanks,
Megan Lawson
Doctoral Candidate
Department of Economics
University of Colorado - Boulder
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
