Relatives' risks: Frailty models of life history data (Q910149)

The article develops a method for applying frailty models based on biological theory of life-history data on related individuals and events. Here a group's frailty is the relative risk (with regard to death, disease, etc.) shared by all members of the group. The main result is the derivation of expressions for the likelihood of kindred survival data and for the distribution of frailty conditional on these data by using integral transforms. The entire history of survival data on grouped individuals or repeated events, with covariates and a mixture of death times and censoring times, can be summarized by three statistics, the total log hazard at observed death times, the total cumulative hazard, and the number of deaths. Hutterite fertility, longevity of twins and the stopping of clinical trials of a new therapy are treated as examples.