# st: time-dependent variables in survival analysis

 From "Stephan Michelson" To Subject st: time-dependent variables in survival analysis Date Fri, 14 Apr 2006 19:47:15 -0700

```Quoting from the manual for Stata 8:

"Enduring variables record characteristics of the subject that endure throughout the time-span, such as sex or
smoking status.  Instantaneous variables are used to describe events that occur at the end of a time-span, such
as failure or censoring." [Survival Analysis at 6]

Yet there is a third kind of variable I do not know how to define in Stata.  That is a description of a
situation or characteristic (not an event) that changes regularly with time.  Age and seniority are two such
variables.  I start with variables for birthdate and hiredate, but if I define a seniority variable sen =
(termination date minus hire date) it will be a scalar.  And as most of the observations have not terminated
(they will be censored), defining such a variable without declaring it to be time dependent will create missing
values.

My question is obvious and simple.  I think it's just that I don't understand the language of stset, because
surely Stata knows how to handle such variables, define them so they take different values on each termination
date.

But *I* don't know how to define them, and I don't find the manual helpful in this regard.  The examples go on
and on about multiple observation data, but I have a single observation per employee, stating when he/she was
born, when hired, when terminated and if so, whether voluntarily.  Where "when terminated" is blank, Stata will
understand that the observation is censored (or I define a binary variable "term" to do so).  I expect Stata
wants me to conceive of sen as an event, and define it in snapspan.  Yet there is no example of "varlist" that
tells me how to do so.

Besides, seniority and age are not events.  They are time-dependent characteristics.  We need better language,
and clearer instructions.  I will be appreciative of an answer no matter how dumb this question appears:  How do
I define time-dependent variables that are not events, but take on different values at the times of events
(ultimately to become independent variables in stcox)?.

Stephan Michelson
Longbranch Research Associates

*
*   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/
```