QIC Program and Model Selection in GEE Analyses

Abstract

The generalized estimating equation (GEE) approach is a widely used statistical method in the analysis of longitudinal data in clinical and epidemiological studies. It is an extension of the generalized linear model (GLM) method to correlated data such that valid standard errors of the parameter estimates can be drawn. Unlike the GLM method, which is based on the maximum likelihood theory for independent observations, the GEE method is based on the quasilikelihood theory and no assumption is made about the distribution of response observations. Therefore, Akaike's information criterion, a widely used method for model selection in GLM, is not applicable to GEE directly. However, Pan (Biometrics 2001; 57: 120-125) proposed a model-selection method for GEE and termed it quasilikelihood under the independence model criterion. This criterion can also be used to select the best-working correlation structure. From Pan's methods, I developed a general Stata program, qic, that accommodates all the distribution and link functions and correlation structures available in Stata version 9. In this paper, I introduce this program and demonstrate how to use it to select the best working correlation structure and the best subset of covariates through two examples in longitudinal studies.

The cost of depression – A cost analysis from a large database

“…By contrast, GEE accounts for this correlation of repeated measurements within an individual person (Liang and Zeger, 1986;Twisk, 2004). Therefore, we selected an independent working-correlation structure via the QIC program in Stata, developed by Cui (2007). Like cost data in general, the dependent variable is not normally distributed and heteroscedastic.…”

Factors associated with professional support access among a prospective cohort of methamphetamine users

“…A multivariate GEE model, achieved by backwards elimination and controlled for age and sex, identified characteristics and behaviours that were independently associated with current service access for methamphetamine use (Table 3). Adequacy of the final multivariate GEE model was assessed using the quasi likelihood under the independence model criterion (Cui, 2007). All data analyses were conducted using Stata Version 11.1 (Statacorp LP, Texas, USA) with a significance level of p b .05.…”

Microclimate and Human Factors in the Divergent Ecology of Aedes aegypti along the Arizona, U.S./Sonora, MX Border

“…aegypti presence or absence, and exchangeable, unstructured, first order autoregressive, stationary, and first order nonstationary correlation structures were compared. All potential fixed effects were included in the GEE, and best fitting correlation structure with the lowest quasi-likelihood under the independence model criterion (QIC) value was preserved (Pan, 2001, Cui, 2007. A stepwise backwards elimination variable selection with a stopping criterion of p < 0.05 produced the final GEE models.…”