570 – Strategies and Examples for Teaching Statistics in Health Science
Teaching of Multiple Regression Should Reflect the Way It Works
David Hoaglin
University of Massachusetts Medical School
Currently multiple linear regression is usually taught in ways that limit students' understanding and lead to mistakes in applications. The most important shortcoming involves the interpretation of regression coefficients, specifically the contribution of the other explanatory variables. Also, when the definitions of the regression coefficients are presented, the role of those other variables is often overlooked. The workings of least-squares regression are straightforward to understand, and students can be given adequate explanations without technical details. The benefits extend to other regression methods, including logistic regression and survival analysis.