Generalized Estimating Equations

The basic modeling tool considered in this chapter remains generalized linear models. The chapter begins with the subject of data layout for a situation involving longitudinal data. The iterative routine for estimation of regression coefficients also provides iterative estimation of correlation with methodology that utilizes and is consistent with the structure that is chosen by the user. The generalized estimating equation methodology provides a consistent estimator of β, even if the correlation structure is incorrect because the procedures use consistent estimates of the variance-covariance structure. The empirical or robust estimator and the model-based estimator give similar results, which implies that the correlation structure assumed is a good practical choice. The chapter discusses the examples regarding these two types of estimators of standard errors. The foundation for the use of generalized estimating equations for nonnormal responses is similar to the one developed for normally distributed responses. Controlled Vocabulary Terms autocorrelation; estimating equations; non-linear iterative partial least squares