A functional method for the conditional logistic regression with errors-in-covariates

Abstract

In this article, we develop a functional approach for handling errors-in-covariates in matched case–control studies which are commonly analysed through the conditional logistic regression. We propose to estimate the parameters from a set of unbiased estimating equations. We require that the moment-generating function of the measurement errors exists. We investigate the asymptotic properties of the estimators. The finite sample performance of the method is judged via simulation studies. The proposed methodology is illustrated by analysing the data from the NIH-AARP Diet and Health study.

Keywords:

• Bootstrap variance
• case–control
• conditional likelihood
• estimating-equation
• moment-generating function
• NIH-AARP Diet and health study

Acknowledgements

The author wishes to thank the NIH-AARP study group for providing the NIH-AARP cohort data. He also thanks the editor, associate editor, and a referee for very constructive suggestions which have led to a much improved manuscript. This research was partially supported by NSF grant SES-0961618.