Multiple Linear Regression and Logistic Regression Analysis Using SAS

This chapter aims to provide readers with the basic concepts and data analysis of multiple regression analysis using SAS. For illustrations of data analysis, we used the Bogalusa Heart Study (BHS), a community-based investigation of the early natural history of cardiovascular diseases. Dr Gerald S. Berenson, a Bogalusa native and cardiologist, founded the BHS in 1972, funded by the National Heart, Lung, and Blood Institute. Major concepts demonstrated in this chapter include the independent effect of variables of interest, linear fit assumption, model fitness evaluation, regression diagnostics, multicollinearity, trend analysis, multiple linear regression models, testing the slope of a regression line, best-fit regression model, binary logistic regression, and multiple logistic regression models. Continuous, categorical, and ordinal variables were used for illustration. Criteria for forward, backward, and stepwise model selection procedures are explained. Interaction or effect modification is explained using examples from the dataset. The odds ratio, 95% confidence interval, and their interpretations are provided in binary logistic regression. SAS codes are provided in all of the methods of data analysis.