2019/04/04 by Pourab Roy, Jason P. Fine, Roy, Pourab +3
Social Sciences · Mathematics · Environmental Science · #Insurance, Mortality, Demography, Risk Management #Statistical Methods and Bayesian Inference #Air Quality and Health Impacts
paper · pdf · doi:10.48550/arxiv.1904.02624
In prevalent cohort studies where subjects are recruited at a cross-section,\nthe time to an event may be subject to length-biased sampling, with the\nobserved data being either the forward recurrence time, or the backward\nrecurrence time, or their sum. In the regression setting, it has been shown\nthat the accelerated failure time model for the underlying event time is\ninvariant under these observed data set-ups and can be fitted using standard\nmethodology for accelerated failure time model estimation, ignoring the\nlength-bias. However, the efficiency of these estimators is unclear, owing to\nthe fact that the observed covariate distribution, which is also length-biased,\nmay contain information about the regression parameter in the accelerated life\nmodel. We demonstrate that if the true covariate distribution is completely\nunspecified, then the naive estimator based on the conditional likelihood given\nthe covariates is fully efficient.\n