Estimating Logistic Regression Parameters for Bivariate Binary Data

SUMMARY Consider bivariate binary data, with possibly different covariates for each marginal binary observation. Suppose that the correlation between paired observations is a nuisance, and the interest lies in estimating logistic regression parameters associated with the covariates. We consider the estimator obtained by using the independence estimating equation, with corrected variance. This estimator can be very efficient and robust compared with maximum likelihood or the estimator of Liang and Zeger. Some comments are made concerning higher dimensional multivariate data.