Simulation studies confirmed the validity of the proposed Wald-type rank test using GEE in large sample studies as well as that performances of the proposed small sample corrected tests are similar to the Wald-type rank test proposed in previous studies in a two-way repeated measures design.
Analysis was conducted using generalized linear models (GLM) with Binomial outcomes and the logit link function with exchangeable autocorrelations within providers using generalized estimating equations (GEE) in using the function 'gee' in the 'gee' package.
We conducted our analysis using the ALR implementation of GEE in SAS V8.02 (SAS Institute Inc., Cary, NC).
The comparison between the three symptom complexes was performed using GEE in which the three complexes were input as repeated measurements within each person with associated indicator predictor variables.
To examine the factors predicting women's actions in helping their husbands quit smoking, we applied a generalised estimating equation (GEE) in an unstructured correlation matrix with binary logistic link function in the analysis.
The two latter models are slightly different from GEE in the method of handling correlation in the data and, as expected, the analysis using these models did not demonstrate any significant differences, indicating the robustness of our models to the method of handling correlation.
