Regression analysis with clustered data.

Abstract

Clustered data are found in many different types of studies, for example, studies involving repeated measures, inter-rater agreement studies, household surveys, crossover designs and community randomized trials. Analyses based on population average and cluster specific models are commonly used for estimating treatment (exposure) effects with clustered data. This paper discusses conditions when one or both types of models are appropriate for estimating causal effects and when there is agreement between population average and cluster specific analyses. Applications of survey sampling methods to the robust estimation of standard errors of estimated treatment parameters are discussed.