# Re: st: -stset- with stock-sampling, frailty and TVC

 From rgutierrez@stata.com (Roberto G. Gutierrez, StataCorp) To statalist@hsphsun2.harvard.edu Subject Re: st: -stset- with stock-sampling, frailty and TVC Date Tue, 07 Jul 2009 08:24:49 -0500

```Antoine Terracol <Antoine.Terracol@univ-paris1.fr> asks:

> Suppose I have a stock-sample of duration data where I happen to know the
> values of the covariates before the first observation of the individuals
> (because of a retrospective questionnaire, for example).

> A typical individual might look like:

> id   x    time  fail  tstock
> 1    1    1.5    0     2
> 1    2    2.5    0     2
> 1    3    4      1     2

> I am interested in estimating a parametric model with frailty.

> What would be the correct way to -stset- my data so that Stata uses the path
> of the covariates before the sampling date?

> if I specify

> . stset time, id(id) fail(fail) enter(tstock)

> Stata will discard the first row of the individual in the above example (_st
> will be 0).

> In a model without frailty, this would not matter since the sub-survival
> functions between 0 and enter() disappear from the likelihood.

> With frailty, however, this is not true since we deal with the ratio of
> unconditional survival functions (integrated w.r.t the distribution of the
> frailty). Using the above -stset- command, we in fact assume that the
> covariate is constant from time 0 to the first observed value with
> date>=enter()

Unfortunately, there is no current -stset- syntax that will incorporate
covariate patterns at times when subjects are not considered "under
observation" into the analysis.  However, there is merit to the idea of being
able to incorporate such retrospecitive information into the analysis.  As
such, it is something we'll look into.

--Bobby
rgutierrez@stata.com
*
*   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/
```