A1179
Title: Adjusted predictions in generalized estimation equations
Authors: Francis Hui - The Australian National University (Australia) [presenting]
Abstract: Generalized estimating equations (GEEs) are a popular regression approach that requires specification of the first two marginal moments of the data, along with a working correlation matrix capturing the covariation between responses, e.g., temporal correlations within clusters in longitudinal data. The majority of research and application of GEEs has focused on the estimation and inference of regression coefficients in the marginal mean. When it comes to prediction using GEEs, practitioners often also base it on the regression model characterizing the marginal mean. The aim is to propose an adjustment to predictions in GEEs based on utilizing information in the assumed working correlation matrix. By viewing the GEE from the perspective of solving a working linear model, ideas are borrowed from universal kriging to construct a predictor that leverages temporal correlations between the new and current observations within the same cluster. Some theoretical conditions are established for the proposed adjusted GEE predictor to outperform the standard unadjusted predictor. Simulations show that even when the working correlation is misspecified, adjusted GEE predictors can achieve better performance relative to standard GEE predictors, the so-called "oracle" GEE predictor using all time points, and potentially even cluster-specific predictions from a generalized linear mixed model.
