Stata Tip 8: Splitting Time-span Records with Categorical Time-varying Covariates

Abstract

In survival analysis, time-varying covariates are often handled by the method of episode splitting. The stsplit command does this procedure very well, especially in the case of continuous time-varying variables such as age or time in study. Quite often, however, we are interested in evaluating the effect of a change in some kind of categorical status or the occurrence of some secondary event. For example, we might be interested in the effect of the birth of a child on the risk of divorce or the effect of having completed further education on the chances of upward occupational mobility.