A GENERALIZED ESTIMATING EQUATIONS APPROACH FOR SPATIALLY CORRELATED BINARY DATA - APPLICATIONS TO THE ANALYSIS OF NEUROIMAGING DATA

This paper proposes a generalized estimating equations approach for th e analysis of spatially correlated binary data when there are large nu mbers of spatially correlated observations on a moderate number of sub jects. This approach is useful when the scientific focus is on modelin g the marginal mean structure. Proper modeling of the spatial correlat ion structure is shown to provide large efficiency gains along with pr ecise standard error estimates for inference on mean structure paramet ers. Generalized estimating equations for estimating the parameters of both the mean and spatial correlation structure are proposed. The use of semivariogram models for parameterizing the correlation structure is discussed, and estimation of the sample semivariogram is proposed a s a technique for choosing parametric models and starting values for g eneralized estimating equations estimation. The methodology is illustr ated with neuroimaging data collected as part of the National Institut e of Neurological Disorders and Stroke (NINDS) Stroke Data Bank. A sim ulation study demonstrates the importance of accurate modeling of the spatial correlation structure in data with large numbers of spatially correlated observations such as those found in neuroimaging studies.