ABSTRACT
As discussed in the previous chapter, the standard strategy in using parametric models of time-dependence is to consider measures of time as proxies for time-varying causal factors that are difficult to observe directly. However, available theory in the social sciences normally provides little or no guidance for choosing one parametric model over another. For example, as shown in the previous chapter, whether job-specific labor force experience changes linearly (Gompertz model) or log-linearly (Weibull model) over time can hardly be decided on a theoretical basis. Thus, it is important to empirically check the adequacy of models upon which inferences are based. One way of doing this was demonstrated in the previous chapter by using likelihood ratio tests as a tool for comparing the improvement in the goodness-of-fit of alternative models. This method is, however, limited to nested models.
