Standard Article

Generalized Estimating Equations


  1. Ross Darnell

Published Online: 15 JAN 2013

DOI: 10.1002/9780470057339.vag007m.pub2

Encyclopedia of Environmetrics

Encyclopedia of Environmetrics

How to Cite

Darnell, R. 2013. Generalized Estimating Equations . Encyclopedia of Environmetrics. 3.

Author Information

  1. CSIRO Mathematics, Informatics and Statistics, Brisbane, Queensland, Australia

Publication History

  1. Published Online: 15 JAN 2013


Interdependence is a characteristic feature of many environmental processes. However, the complexities for statistical analysis introduced by such interdependence have hampered the progress of environmental science. Why this is so can be understood by considering the workhorse of standard statistical analysis, the generalized linear model (GLM). A typical GLM divides variation in some measure into a systematic part and a random part. Typically substantive theory guides the construction of the model for the systematic part. For the random part, although some guidance may be obtained from knowledge of the distributional consequences of a simple and plausible stochastic process that might be responsible for this variation, and from experience gained with other comparable data, we rarely have complete confidence in our choice of distribution. In the case of spatial, longitudinal and multivariate outcomes the choice is among multivariate distributions since we need to describe the association across outcomes as well as within outcome variability. This becomes especially difficult for problems where the multivariate normal distribution is not a suitable choice for the random part of the model, since the available alternative distributions lack flexibility. Under these circumstances it is often helpful to use generalized estimating equations (GEEs). The GEE approach, while making weaker distributional assumptions than that required for a fully parametric linear model, maintains the properties of consistency and asymptotic normality of parameter estimates.