Bias-Corrected Inference for the Conditional Logistic Regression

• Activity Number: 315
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jenny Sun*+
• Companies: Texas A&M University
• Keywords: Bias ; Conditional Logistic Regression ; Estimating Equation ; Jackknife Method ; Matched Case-Control Study

Abstract

The conditional maximum likelihood estimator (CMLE) of the log-odds-ratio for matched case-control study is biased even for moderate sample sizes. Most existing methods for bias correction subtract the bias from CMLE. The CMLE of Jackknife method is computed by deleting one stratum at a time from the original data set. However, the CMLE could produce infinite value or may not exist for a separable data set. By modifying the conditional logistic likelihood score equation, we propose a new method to estimate the parameters that corrects the first-order bias. Its asymptotic properties are examined. A closed form expression is also derived for a dichotomous exposure variable. Finite sample performance of the estimator is investigated via a simulation study. The method is applied to a matched case-control data set from a low-birth-weight study.