Generalized Estimating Equations

Written by Scott Martin
Bookmark and Share

If you are working with panel data and longitudinal analysis and need to manage the correlation structure, you should consider using generalized estimating equations (GEE). In fact, generalized estimating equations are one of the most important ways to analyze correlated data. This data is often encountered when the same units of data are collected across consecutive points over time.

Beginnings of Generalized Estimating Equations

Generalized estimating equations were first described in the 1986 publication by Liang and Zeger as a way to analyze correlated data. They were an outgrowth of generalized linear models and are one of the models used to model panel data. One of the ways to work with this data is through a random effects model.

The generalized estimating equations methodology is more frequently used when the dependent variable is dichotomous rather than continuous. Furthermore, one of the main focal points of GEE is on the specification of the model and the hypothesis testing.

There are certain parameters that need to be defined when working with a regression model using generalized estimation equations. The first item that needs to be defined is the dependent variable distribution. Additionally, the link function, the independent variables, and the covariance of the repeated measurements need to be specified.


Bookmark and Share