Common Errors in Statistics (and How to Avoid Them)
This book may be profitably used by scientists, physicians, lawyers, business types and students
It is always a pleasure to review a book on statistics (bias intended)! Especially one that is well-written, compact and well-constructed for its intended audience. Common Errors in Statistics (and How to Avoid Them), 4th Edition, is written for the non-statistician and can be readily assimilated by undergraduate students with a single statistics class under their belt. The authors state that the purpose of the book is to “provide a mathematically rigorous but readily understandable foundation for statistical procedures.” From the standpoint of a statistician, I found the mathematical rigor to be of a very low level (for the rigor, consult some of the many references that they cite) but the understandability very high.
They make liberal use of reports from the statistical literature and draw on papers from a wide variety of disciplines for both uses and misuses of statistics. As in formal courses, if all they dwelt on were a menu of tests and their correct uses, the book would be not only boring, but incomplete. By highlighting misuses from everyday examples, they make the text more interesting as well as adept at drilling home the important points.
For an overview, peruse the Table of Contents:
Part I Foundations
1. Sources and Error
2. Hypotheses: The Why of Your Research
3. Collecting Data
Part II Statistical Analysis
4. Data Quality Assessment
6. Testing Hypotheses: Choosing a Test Statistic
7. Strengths and Limitations of some Miscellaneous Statistical Procedures
8. Reporting Your Results
9. Interpreting Reports
Part III Building a Model
11. Univariate Regression
12. Alternate Methods of
13. Multivariable Regression
14. Modeling Counts and Correlated Data
The organization is quite intuitive and logical, while not burying the reader in the minutiae of theory, and serves well to support the pedagogical value of the text. In some modern textbooks (mercifully very few) it is maddening to find salient points buried in paragraphs as verbiage. The authors wisely set off lists from the text, bolded and sequentially numbered. Recognizing that pictures are oft times better than words, the value of graphics and exploratory data analysis is stressed. Up front design and a thorough knowledge of the problem and variables to take into account are highlighted, and I cannot see their caveat concerning “letting the computer do your thinking” enough.
They often do the reader a service by suggesting modern software, holding the heavy-duty math to a minimum, and repeating the more important lessons. The inclusion of many tips and guidelines where no hard and fast rules exist is also very helpful to the beginner. All-important experimental design is well-supported by the sections on statistical power and sample size, misuse of which will sink many an experiment. What is particularly pleasing is the handbook-like style where, after a test is introduced, the authors list the strong points and limitations. Although not intended to be an exhaustive text, it introduces the novice to the basics of what is to be done in most situations. (After that, the beginner is encouraged to see a statistician).
The limitations in the book are few: the use of color — especially in highlighting the most important points — would be helpful, and there is the occasional confusing paragraph, for example under the Neyman-Pearson Theory section of Chapter 2 where the null/alternate hypothesis is introduced verbally in one form, then inverted in the following table. Also, in most of the tables, the shades of grey used many times obscure the print.
The few negatives are far outweighed by the positives, however, and this book may be profitably used by scientists, physicians, lawyers, business types and students. Recommended.
Common Errors in Statistics (and How to Avoid Them), 4th Ed., by Phillip I. Good and James W. Hardin. John Wiley & Sons, Hoboken, N.J. (2012). Pb, 336 + xiv. $59.95
John Wass is a statistician based in Chicago, IL. He may be reached at editor@ScientificComputing.com.