Applied Regression Including Computing and Graphics, by R. D. Cook and Sanford Weisberg, Copyright 1999, John Wiley and Sons, ISBN 0471-31711-X. Reprinted by permission.
Go back to index.
This textbook is about regression, the study of how a response variable depends on one or more predictors. Regression is one of the fundamental tools of data analysis. This book is intended for anyone interested in applying regression to data. The mathematical level required is intentionally low; prerequisite is only one semester of basic statistical methods.
The main features of the book are:
We have used drafts of this textbook for several years as a basis for two different courses in applied regression analysis, one intended primarily for advanced undergraduate and beginning graduate students in fields other than statistics, and the other for first-year statistics graduate students. Others have also used drafts for students in social science, business and other disciplines. After completing these courses, our students can analyze regression problems with greater ease and more depth than could our students before we began to use this book.
Integrated into the text is discussion of a computer package called Arc. This is a user-friendly program designed specifically for use while studying this material, and for applying the ideas learned to other data sets. The program permits the user to do the analyses discussed in the book easily and quickly. It can be downloaded for free from the Internet site for this book at the address given in the Appendix; versions are available for Windows, Macintosh and Unix.
For those readers who prefer to use other statistical packages, we include on our Internet site descriptions of how a few of the major packages can be used for some of the calculations.
When writing the book, we envisioned the reader sitting at a computer and reworking the examples in the text. Indeed, some of the text can be read as if we were next to the reader, suggesting what to do next. To maintain this low-key style, we have tried to avoid heavy algebra. References and technical comments are collected in the Complements section at the end of each chapter.
The book is divided into four parts:I: Introduction. The first part consists of five chapters that present many basic ideas of regression to set the stage for later developments, including the one-sample problem, using and interpreting histograms and scatterplots, smoothing including density estimates and scatterplot smoothers, and bivariate distributions like the normal. These chapters weave together standard and non-standard topics that provide a basis for understanding regression. For example, smoothing is presented before simple linear regression. II: Multiple Linear Regression. The second part of the book weaves standard linear model ideas, starting with simple regression, with the basics of graphics, including 2D and 3D scatterplots and scatterplot matrices. All these graph types are used repeatedly throughout the rest of the book. Emphasis is split between basic results assuming the multiple linear regression model holds, graphical ideas, the use of transformations, and in the last few chapters of this part the use of graphs as the basis for diagnostic methods. III: Graphics. The third part of the book, which is unique to this work, shows how graphs can be used to better understand regression problems in which no model is available, or else an appropriate model is in doubt. These methods allow the analyst to see appropriate answers, and consequently have increased faith that the data sustain any models that are developed. IV: Other models. The last part of the book gives the fundamentals of fitting generalized linear models. Most of the graphical methods included here are also unique to this book.
For a one-quarter (30 lecture) course, we recommend a very quick tour of Part I (3-5 lectures), followed by about 18 lectures on Part II, with the remainder of the course spent in Part III. For a semester course, the introduction can be expanded, and presentation of Parts II and III can be slowed down as well. We have used Part IV of the book in the second quarter of a two-quarter course; completing the whole book in one semester would require a fairly rapid pace. Some teachers may prefer substituting Part IV for Part III: we don't recommend this because we believe that the methodologies described in Part III are too useful to be skipped.
Some of the material relating to the construction of graphical displays is suitable for presentation in a laboratory or recitation session by a teaching assistant. For example, we have covered Section 7.1 and most of Chapters 5 and 8 in this way. Further suggestions on teaching from this book are available in a teacher's manual; see the Internet site for more information.
On the Internet site for this book, we provide additional references for reading for those who would like a more theoretical approach to this area.
Throughout this work, we have been supported by the National Science Foundation's Division of Undergraduate Education. Without their support, this book would not exist.
R. D. Cook & Sanford Weisberg
St. Paul, Minnesota
Go back to index.