|
Abstract:
|
In introductory statistics courses, students are often warned about "confounding variables." But this warning rarely comes along with a serious treatment. For instance, there is rarely any quantitative consideration of the effect that covariates can have and the techniques for dealing with them. They are merely "lurking" in the background. Much of what goes on in experimental design or the analysis of observational data stems from the importance of covariates. In this presentation, I'll show some techniques for teaching about covariates in a serious way with introductory students. The approach I take is a substantial departure from the usual introductory material: introduction of ideas from statistical modeling, use of simulations, graphical representations of the process of fitting models to data. By taking confounding seriously, students can reach a much better understanding.
|
