I'm interested in testing linear contrasts between coefficients for three
treatment groups, Groups 1, 2 and 3 after Cox proportional hazards regression.
For example, as part of a set of orthogonal contrasts, I would like to compare
Group 3 to the average of Groups 1 and 2.
For linear regression, if trt is the treatment variable, and dep the dependent
variable, I could -xi: regress dep i.trt-, and then -lincom 0.5 * _b[_cons] +
0.5 * (_b[_cons] + _b[_Itrt_2]) - (_b[_cons] + _b[_Itrt_3])- to test the linear
combination of coefficients, (0.5 * Group 1) + (0.5 * Group 2) - (1 * Group 3)
= 0. -stcox- doesn't provide a _cons term for this -lincom- syntax.
Stata reports that, with collection of terms, this linear combination becomes
.5 _Itrt_2 - _Itrt_3 = 0. Question: am I therefore testing the analogous
hypothesis after -xi: stcox i.trt- when I enter -lincom 0.5 * _b[_Itrt_2] -
_b[_Itrt_3]- (with or without the -hr- option)?
-stcox- allows -lincom-, but should I be thinking of the contrast in terms of
-nlcom-? There are no time-varying covariates in this particular model, and
the dataset has been -stset- appropriately.
Joseph Coveney
*
* 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/
