Relative mortality for correlated lifetime data.
Song-JX; Wassell-JT; Kapadia-A
Comput Stat Data Anal 2004 May; 45(4):849-864
Comparing correlated lifetimes for a group of individuals to a standard reference population may require variance adjustment of marginal model estimates or the use of conditional random effects models with shared frailty. We present the cumulative relative mortality, the marginal model robust variance estimator and the frailty models in estimating relative mortality for individuals who have correlated lifetimes due to group clustering. The positive stable and gamma frailty distributions are considered in the frailty models. The performances of both marginal and frailty models are compared using simulation. Applying these methods, we show siblings for centenarians had longer life spans than their US cohort reference population.
Mortality-data; Mathematical-models; Survival-rate; Lifespan; Computer-models; Data-processing; Epidemiology; Risk-analysis; Risk-factors; Statistical-analysis;
Author Keywords: correlated survival data; cumulative relative mortality; marginal survival mode; frailty model; robust variance estimator; taylor linearization
James X. Song, Bayer Pharmaceuticals Corporation, Department of Biometry, 400 Morgan Lane, West Haven, CT 06516
Computational Statistics & Data Analysis