Issues in smoothing empirical hazard rates.

"This chapter presents a smooth estimator of the hazard rate using a variable-span running loglinear specification that allows investigators to maintain only mild assumptions about the functional forms of population heterogeneity and time inhomogeneity in the rate. This estimator is useful both in exploratory data analyses (EDA) and in checking parametric assumptions. Examples drawn from data on the transition to first marriage for women in the U.S. show that several common parametric assumptions are violated in these data and illustrate that the smoothed hazard estimator can yield important insights not easily obtained from more conventional methods."