Visualizing Categorical Data

In the report on the second edition (Ziegel 1992), I encouraged anyone who did not own the first edition (1E) to “call SAS and order this new edition right away.” I also noted that “if you have the First Edition, I don’t know that you need to buy this new one.” Now it has been 10 years since the 2E. The 1E was published in 1986. I will reiterate both comments. Any statistician or SAS user who does regression analysis must own this book. There also is still not a lot of additional material relative to the 1E or the 2E that would encourage the purchase of an updated version for your personal library. In part there is not a lot in classical regression methodology that is new in SAS, or elsewhere for that matter. In addition, there is a large companion volume (Allison 1999, reported by Ziegel 2000 for Technometrics), that presents the methodology for the one area in regression analysis that has received a lot of attention in the past 10 years. Another popular new regression procedure, at least in the world of chemistry, PROC PLS, is not discussed at all in the book. All of the SAS procedures that do regression analysis in some way are summarized in Chapter 1, but there is no explanation about what will or will not be discussed in the book. The basic chapter on PROC REG, Chapter 2, adds a new section on the datasets that are created by the OUTEST= and TABLEOUT options. There is a brief note about ODS output. Some of the other chapters also comment on ODS output. Chapter 5 was renamed from “Polynomial Regression” to “Curve Fitting” to allow the inclusion of material on nonparametric regression and PROC LOESS. Chapter 6, “Special Applications of Linear Models,” has a new section for analyzing binary response variables using PROC LOGISTIC. There is a brand new Chapter 8, “Using SAS/INSIGHT Software for Regression,” that goes through all the different regression capabilities of PROC INSIGHT. See Ziegel (1992) and the review for the 1E, Mason (1988), for more information on the content of this book. Two things that I would have liked to have seen, in addition to discussion of PROC PLS, are more use of SAS/GRAPH for graphics and inclusion of SAS macros for some of the many other regression methodologies, like MARS, for which there are no procedures. The authors eliminated the appendixes in the 2E that had the code for the examples. One can always get that stuff from Web sites. Any further contemplation of the incompleteness of the book as a regression textbook simply reflects the aspects of the methodology that are absent as procedures in SAS.