Workshop on Regression Graphics

R. Dennis Cook and Sanford Weisberg, University of Minnesota


This course will provide an overview of the ideas of regression graphics, which is a collection of methodological tools that are used to discover and understand dependence of a response on predictors primarily through the use of simple graphs in one, two and three dimensions. It will mostly follow the outline of the book An Introduction to Regression Graphics, published by Wiley in 1994.

Prerequisite for this course is familiarity with standard regression methodology at the level of one of the major textbooks in this area. The course will also introduce the R-code, which is Xlisp-Stat computer code that accompanies the book and allows the user to use all the methods described in the book and in the course. The program is very easy to use, and requires no knowledge of lisp.


The objectives of this course is to help university-level instructors gain the confidence to use regression graphics in their own teaching, and to provide training in the methodology for practicing statisticians for use in their own work. Particpants will also be given access to the R-code2, the lastest version of the R-code.

Outline of the course

  1. Introduction illustrating graphical issues.
  2. Foundations: Structural dimension, sufficient summary plots, linear predictors, first graphical inference
  3. Finding summary plots and exploring the importance of linear predictors.
  4. Illustrations, 3D plots, residual plots
  5. Graphics for regressions with a binary response
  6. Graphics for model assessment.
Throughout the course, we will include hints for teaching, and demonstrate the use of the R-code for doing the analyses suggested.

Target population

The target population is graduate students, university-level instructors who teach regression analysis, and practicing statisticians who wish to learn the latest ideas in using graphs to understand regression problems.

Learning outcomes

Regression analysis is one of the fundamental tools of the practicing statistician. The traditional role of graphics in regression, at least with many predictors, has really been peripheral, dealing mostly with questions of model adequacy. Regression graphics moves graphs to the center of analysis. This requires some new theory, but this theory can be presented at a very general and intuitive level. Our hope is to encourage the participants to use this approach in their own work, and particularly in their own teaching.