ABSTRACT
The baseline rates for the multiple events may all be different, or may be restricted to be common for events of the same type. The notation for the developments is somewhat awkward. This chapter describes the discrete special case of a three-dimensional version of the Dabrowska representation is used to develop a nonparametric estimator of the trivariate survivor function, and asymptotic distribution theory is sketched. It outlines the copula approach to trivariate failure time regression. The chapter discusses rather flexible trivariate regression approach that involves the modeling and estimations of marginal single, double and triple failure hazard rates. It considers application to the Women’s Health Initiative’s postmenopausal hormone therapy trials. Expression breaks the marginal probabilities and pairwise and trivariate dependency ratios into product integrals for ease of estimation.
