Premium
Cox regression in cohort studies with validation sampling
Author(s) -
Chen YiHau
Publication year - 2002
Publication title -
journal of the royal statistical society: series b (statistical methodology)
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.523
H-Index - 137
eISSN - 1467-9868
pISSN - 1369-7412
DOI - 10.1111/1467-9868.00324
Subject(s) - covariate , estimator , statistics , proportional hazards model , cohort , simple random sample , stratified sampling , mathematics , regression analysis , sampling (signal processing) , sample (material) , set (abstract data type) , sample size determination , regression , econometrics , computer science , medicine , population , chemistry , environmental health , filter (signal processing) , chromatography , computer vision , programming language
An estimation procedure is proposed for the Cox model in cohort studies with validation sampling, where crude covariate information is observed for the full cohort and true covariate information is collected on a validation set sampled randomly from the full cohort. The method proposed makes use of the partial information from data that are available on the entire cohort by fitting a working Cox model relating crude covariates to the failure time. The resulting estimator is consistent regardless of the specification of the working model and is asymptotically more efficient than the validation‐set‐only estimator. Approximate asymptotic relative efficiencies with respect to some alternative methods are derived under a simple scenario and further studied numerically. The finite sample performance is investigated and compared with alternative methods via simulation studies. A similar procedure also works for the case where the validation set is a stratified random sample from the cohort.