Quantitative Genetic Epidemiology

Introduction

Mission

Members

Software Downloads

Papers

Introducing the QGE EE Software Package

Outline: Features

Components and Options

Compatibilities

How to get EE

EE, which stands for estimating equations, is a software package which is useful for regression analysis of univariate or multivariate responses with multiple covariates. Regression analysis is one of the most commonly used statistical techniques because a vast number of different application can be structured and analyzed as a regression problem, such as simple comparisons of means between two groups by t-test or between multiple groups by analysis of variance, somewhat more complex problems arising from longitudinal studies, and clustered sampling survey studies or family studies. This generality is the essential motivation for developing EE as a general package of regression analysis.

Traditionally, regression techniques have also been categorized by the specific link functions that depict the relationship between outcomes and covariates. For example, linear regression assumes that a linear combination of covariates is linearly associated with or determines the average of the outcome. Logistic regression assumes that the average of the outcome is the logit of a linear combination of covariates, and is often used for analyzing binary outcomes.

Features

While many regression techniques, and hence relevant computer packages, have been developed for regression analysis, EE has a couple of unique features.

First of all, EE is developed based on an estimating equation technique, which differs from the method of maximum likelihood, or quasi-likelihood. The method of estimating equations requires only assumptions on relevant moments rather than knowledge of the entire distribution required for constructing likelihood functions or unnecessary assumptions of moments in constructing quasi-likelihood. Hence, the estimates obtained by this technique are valid under much more general condition than those by the other techniques, i.e., they are more robust.
Secondly, EE permits simultaneous regression analysis of means, variances, and covariances on covariates, the generality of which sets it apart from two other programs, GEE and GEE2, currently available.

Components and Options

EE provides several options for performing univariate and multivariate analysis. Specifically, when a single response is identified, EE automatically executes the algorithm for univariate analysis. In performing the univariate analysis, one has the option of regressing only the mean of the response on covariates or both the mean and variance of the response on covariates, with a choice of a number of link functions, including linear, logistic and exponential. When a vector of multivariate responses is supplied, EE automatically executes the algorithm for multivariate analysis. In the multivariate analysis, one has the option of focusing on the marginal means and covariances of the outcomes. In either cases, one has a choice of link functions for the marginal means, variances and covariances, as well as a choice of covariates to be included in their respective mean link functions. The flexible choice of covariates into these first two moments allows the user to explore problems ranging from longitudinal studies, studies with repeated measurements, or studies with multiple outcomes.

Compatibilities

EE is available for IBM compatible PCs running MS Windows. The MS Windows version is provided with an easy to use graphical interface. Click here to view a screen shot of the EE Model Specification Window.

How to get EE

EE is available by http from our server.