The use of twin and family survival data in the population studies of aging: statistical methods based on multivariate survival models
86 pages. Odense, University of Southern Denmark (2001)
How do genes and environment influence human aging and survival? Traditionally such questions are addressed by analyzing data on identical and fraternal twins using models of quantitative genetics. However, survival and health history data are often incomplete (because of censoring) and subject to selective sampling due to the limitations of population registries. Unfortunately, traditional methods of twin analysis cannot efficiently deal with these problems. The aim of this project is to develop statistical methods which combine techniques of multivariate survival analysis with methods of quantitative genetics which would allow for genetic analysis of health history data in the presence of censoring and truncation. The approach is based on multivariate frailty models in which the primary variable of interest is the age-specific susceptibility to disease or death. We investigate the questions of identifiability of frailty models and develop a general framework for parameter estimation by maximum likelihood and the EM-algorithm based on the multivariate Laplace transform of the frailty distribution. The two-level model construction allows for genetic analysis of the frailty variable including methods of quantitative genetics, segregation and genetic linkage analysis. We also consider generalizations of the frailty concept such as conditional Markov frailty models and stochastically changing frailty models. The approach is illustrated by the analysis of survival data on Danish, Swedish and Finnish twins and by a number of simulation studies.