|Speaker: Stephen P. Jenkins, University of Essex|
This talk provides a brief overview of discrete time (grouped data) proportional hazards regression models and introduces the author’s pgmhaz program for estimating such models. A notable feature of the program is the ability to estimate models incorporating “frailty” (“unobserved heterogeneity” to economists) using a Gamma mixture specification following (Meyer, Econometrica 1990). Other notable features are the flexibility in baseline hazard specifications (from fully parametric to parametric, as specified by the user), and ability to use time-varying covariates. Various Stata programming issues may also be discussed (pgmhaz uses ml method deriv0). The talk draws on the author’s article in STB-39 but, time permitting, will illustrate the model using unemployment insurance benefit spell duration data for cohorts of Spanish men.