[ LOGLINK ] [ OPTIONS ] [ OUTPUT_GROUPS ] [ STATEMENTS ]
The LOGLINK procedure in SUDAAN fits log-linear regression models to cluster-correlated count data not in the form of proportions. The counts are typically counts of events in a Poisson-like process. The LOGLINK procedure estimates model parameters using generalized estimating equations (GEE). LOGLINK implements two robust variance estimation methods described in Binder (1983) and Zeger and Liang (1986), as well as a model-based (naive) variance estimation method. A choice of independent vs. exchangeable "working" correlations is also provided. You can specify tests for linear combinations of the model parameters, and you can output many statistics for further hypothesis testing. Also, you can estimate and test linear combinations of the conditional and predicted marginals (generalizations of adjusted group means to non-linear models). Like all of SUDAAN's procedures, LOGLINK is designed to analyze data from complex sample surveys (weighted, stratified, cluster-correlated data) as well as from randomized experiments and other observational studies involving cluster-correlated or longitudinal responses.