Premium
Reweighting estimators for Cox regression with missing covariate data: Analysis of insulin resistance and risk of stroke in the Northern Manhattan Study
Author(s) -
Xu Qiang,
Paik Myunghee Cho,
Rundek Tatjana,
Elkind Mitchell S. V.,
Sacco Ralph L.
Publication year - 2011
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.4380
Subject(s) - covariate , jackknife resampling , missing data , estimator , statistics , regression analysis , econometrics , proportional hazards model , regression , data set , variance (accounting) , computer science , mathematics , accounting , business
Incomplete covariates often obscure analysis results from a Cox regression. In an analysis of the Northern Manhattan Study (NOMAS) to determine the influence of insulin resistance on the incidence of stroke in nondiabetic individuals, insulin level is unknown for 34.1% of the subjects. The available data suggest that the missingness mechanism depends on outcome variables, which may generate biases in estimating the parameters of interest if only using the complete observations. This article aimed to introduce practical strategies to analyze the NOMAS data and present sensitivity analyses by using the reweighting method in standard statistical packages. When the data set structure is in counting process style, the reweighting estimates can be obtained by built‐in procedures with variance estimated by the jackknife method. Simulation results indicate that the jackknife variance estimate provides reasonable coverage probability in moderate sample sizes. We subsequently conducted sensitivity analyses for the NOMAS data, showing that the risk estimates are robust to a variety of missingness mechanisms. At the end of this article, we present the core SAS and R programs used in the analysis. Copyright © 2011 John Wiley & Sons, Ltd.