LOGITSE: A SAS Macro for Logistic Regression Modeling in Complex Surveys

Traditional formulae for standard errors and subsequent statistical significance tests implemented in various popular statistical packages are based on the premise that the data are a simple random sample (SRS) of observations from a superpopulation. Equivalently, the observations are assumed to be independent and identically distributed (IID). For complex analytic surveys, these assumptions are almost always invalid leading to potentially incorrect inferences due to the failure to adjust relevant standard errors of the parameters. Here, we concentrate on binary responses where a logistic regression analysis would be meaningful and introduce a SAS ® macro, LOGITSE, that takes the cluster-correlated nature of the complex survey design into account, thus providing for correct inference.