Teaching regression with simulation

Computer simulations can be used to teach complicated statistical concepts in linear regression more quickly and effectively than traditional lecture alone. In introductory applied statistics classes, the coverage of important statistical topics, such as the nature of the sampling distribution of the simple linear regression slope, the problem of multicollinearity in multiple linear regression, or the danger of extrapolation in predictions from multiple linear regression models, may be shortened or eliminated entirely for lack of time. Simulation can provide a way to introduce these topics in a brief, but memorable, way for introductory students or as the first step in a more thorough treatment for higher level students. This paper describes each simulation, discusses its pedagogical advantages, and gives sample computer output.