(a) Using a 50\% random sample of the McCall data set estimate the KaplanMeier nonparametric survival and integrated hazard function estimates by type of censoring, that is, by whether transition is to full-time or part-time employment. Does the survival function look significantly different for the two groups?

(b) Ignoring the censoring variable for type of spell termination, estimate the hazard model for unemployment duration under the following parametric distributional assumptions: (i) exponential, (ii) Weibull, (iii) log-logistic, and (iv) Cox $\mathrm{P}\mathrm{H}$. Use the same covariates as in this chapter.

(c) Compare models (i)-(iii) and discuss which one you think provides the best fit to the data. What does each model imply regarding the duration dependence (shape of the hazard function) of a spell of unemployment?