Event-history Analysis: Selectivity

Abstract

The purpose of the analysis of event history data is to explain why certain individuals are at a higher risk of experiencing the event(s) of interest than others. This can be accomplished by using special types of methods which are called hazard models. These are log-linear regression models in which the risk of experiencing an event within a short time interval is explained by a set of (time-varying) covariates. In the context of the analysis of event history data, the problem of unobserved heterogeneity, or the bias caused by not being able to include particular important explanatory variables in the regression model, has received a great deal of attention. This is not surprising because this phenomenon, which is also referred to as selectivity or frailty, may have a much larger impact in hazard models than in other types of regression models: unobserved heterogeneity may introduce, among other things, downwards bias in the time effects, spurious effects of time-varying covariates, spurious time-covariate interaction effects, as well as dependence between competing risks and repeatable events. Several model-based approaches have been proposed to correct for unobserved heterogeneity.