Multilevel Analysis of Ordinal Outcomes Related to Survival Data
Application of these continuous-time models to grouped or discrete-time survival data is generally not recommended because of the large number of ties that result. Instead, models specifically developed for grouped or discrete-time survival data have been proposed. Both Han and Hausman (1990) and Scheike and Jensen (1997) have described proportional hazards models incorporating a log-gamma distribution specification of heterogeneity. Also, Ten Have (1996) developed a discrete-time proportional hazards survival model incorporating a log-gamma random effects distribution, additionally allowing for ordinal survival and failure categories. Ten Have and Uttal (1994) used Gibbs sampling to fit continuation ratio logit models with multiple normally distributed random effects. In terms of a GEE approach, Guo and Lin (1994) have developed a multivariate model for groupedtime survival data.