ABSTRACT
Previously, we stressed that time-dependence can be interpreted from various angles. In chapter 7, we discussed time-dependence from a substantive perspective, where it was derived from a theoretically supposed underlying diffusion process, or as an expression of a theoretically important latent causal factor operating in time. In this chapter we approach time-dependence from a methodological point of view and consider it to be a consequence of unobserved heterogeneity. We will see that such a switch in perspective means that scientists can investigate the same data using different assumptions and logically reach valid but perhaps contradictory substantive conclusions (Manski 1993). Thus, distinguishing between competing (substantive and methodological) interpretations of the same data can be considered one of the most challenging problems in causal analysis. Therefore, it is not surprising that the estimation of event history models in the presence of omitted variables has become a prime focus of much technical work in recent years.
