A. A. Siyam <A.A.Siyam@lse.ac.uk> asks:
> A colleague of mine is asking about how to best interpret the "theta" value
> following the estimation of a "gompertzhet" model.
> As an example, we have the following model:
> xi:gompertzhet time i.sex i.tenureg i.newcohort i.marstat i.nmm i.socg
> i.educ i.bathg i.carg i.urind i.pob, dead(outcome) frailty(gamma) robust hr
[output omitted]
> We found a great deal of help in understanding the gompertz model from the
> book "Introduction to Survival Analysis Using Stata" by Cleves,Gould and
> Gutierrez. But, there seems to be very little explanation of how to use and
> interpret the "theta" coeffi cient estimated by the model.
First of all, the equivalent way to fit your model using the more modern
-streg- is
. stset time, failure(outcome)
. xi: streg i.sex i.tenureg i.newcohort i.marstat i.nmm i.socg /*
*/ i.educ i.bathg i.carg i.urind i.pob, dist(gompertz) /*
*/ frailty(gamma) robust
That being said, you'll find the discussion of the frailty variance, theta,
estimated by -streg, frailty()- in Section 15.3 of the aforementioned text
applicable.
Namely, theta is the estimated frailty variance. If significantly different
from zero (in the case of robust no formal likelihood-ratio test is available,
so merely compare the parameter estimate to its standard error), this
indicates significant heterogeneity. This heterogeneity can be thought of as
arising from some unobserved observation-specific effect, or frailty. Theta
measures the variability of these frailties. If close to zero, then the
frailties are virtually non-existent, and you can get by with the standard
Gompertz model.
--Bobby
rgutierrez@stata.com
*
* 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/