A SAS Macro for Hosmer and Lemeshow’s Purposeful Selection Model Building Algorithm: Description and Performance

A common problem in many model-building situations is to choose from a large set of covariates that should be included in the “best” model. An additional consideration in modeling epidemiological data is the inclusion of confounders, which adds a quirk in the modeling procedure in that statistical significance is not the main criteria for keeping predictors in a model. Hosmer and Lemeshow (2000) describe a purposeful selection of covariates algorithm within which an analyst makes a variable selection decision at each step of the modeling process with a primary goal of inclusion of important confounder terms. In this poster we introduce a macro, %PurposefulSelection, which automates this process. We conduct a simulation study to compare the performance of this algorithm with three well documented variable selection procedures in SAS PROC LOGISTIC, FORWARD, BACKWARD, and STEPWISE. Results and implications are discussed in this presentation.