Finite sample adjustments in estimating equations and covariance estimators for intracluster correlations

Bias-corrected covariance estimators are introduced in the context of an estimating equations approach for intracluster correlations among binary outcomes. Simulation study results show that the bias-corrected covariance estimators perform better than uncorrected sandwich estimators in terms of bias and coverage probabilities. Additionally, introduction of a matrix-based bias-correction into the estimating equations considerably improves point and interval estimation for the intracluster correlations. The methods are illustrated using data from a nested cross-sectional cluster trial on reducing underage drinking.