ABSTRACT

This chapter introduces the reader to regression topics, starting with simple linear regression models used to predict the change in one continuous outcome from one predictor. The assumptions of a linear regression model (independence, normality, linearity, and homoscedasticity), calculation of regression coefficients, and inference for predicted values are discussed. Readers will also learn how to evaluate the model fit for linear regression models and will be exposed to a discussion on residual plots and the coefficient of determination (R2). The second half of Chapter 13 notes multiple linear regression topics, such as indicator variables, model selection, interaction terms, regression model diagnostics, and collinearity. Concepts are placed in an epidemiologic framework with a brief discussion on confounding and moderation (effector modification) in regression models. The method of least squares estimation is discussed, and the use of statistical software (SAS and Stata) to develop regression models is demonstrated. Emphasis is placed on reading output from statistical software and interpreting the results in context.