Statistics As Principled Argument

Abelson's Statistics as Principled Argument is a must read for advanced undergraduates or graduate students learning statistics.

Abelson really explains why we do statistics the way we do, and how to construct and deconstruct a statistical argument. Abelson has a very lucid writing style which makes the book a very quick read.

I recommend assigning this book in your statistics courses. It will really help new researchers understand how and why statistics is the language of science.

The author is a psychologist with strong graduate training in mathematics and statistics. He did his graduate work at Princeton and as his dedication indicates. he was very much influenced by the work of John Tukey (who he probably took courses from at Princeton in addition to later interactions). Abelson is a Fellow of the American Statistical Association and is very knowledgeable on the statistical literature from the 1940s up to the present.
The book is mostly expository with a lot of dialogue and interesting applications. Applications are concentrated in psychology, a field that Abelson is most familiar with but coverage is not restricted to psychology. He covers examples from sports, gambling and medicine as well and some of these examples are very good at making points about common misunderstandings a perceptions about probability and statistics.

A main theme is the importance of a reasoned or principled argument in the presentation of results from a statistical analysis. Abelson points out that many researchers are afraid of mathematics and statistics and use it only as a necessary tool in a research project. They want to find a method for turnng the crank and reaching a conclusion. Abelson recognizes that this causes trouble.

There is too much reliance on the 0.05 significance level. Researchers fail to understand the use of a null hypothesis or the fact that conclusion of hypothesis tests are not black and white and have possible errors associated with them. Most researchers do not understand the subtleties of the interpretation of p-values or confidence intervals and many do not know or understand the difference between the frequentist and Bayesian approaches.

Abelson is careful to articulate all of this in a way that statisticians would be proud of and hopefully it will be understandable to researchers as well.

He also spends time in the text describing the counterintuitive streaky nature of random sequences. He refers to this by saying "Chance is lumpy." He illustrate this with examples from ESP experimentation and the study of the hot hand in basketball. This is all time well spent. Abelson also points out the gambler's fallacy of believing that a batter is "due" for a hit if he has not gotten one for many at bats or that red should come up on the next spin of the roulette wheel after a string of 5 or 6 blacks.

Equivalence and multiple comparisons are two topics that are well-covered in the book. The author also speaks well for Tukey's exploratory approach to data analysis and includes some graphics including stem-and-leaf plots. But aside from a few graphs and tables, there is very little mathematics and no formulas or derivations. This can make it a little difficult for the mathematical statisticians at times. Yet this may be viewed as a blessing by the less mathematically inclined researcher.

In any case it is worth reading for anyone involved in statistical analysis especially graduate students and researchers.

This is the most informative book on statistics that I have ever read. In every chapter there is something to stimulate thinking. I only wish someone had recommended it to me last decade rather than last year.

Anyone who has struggled to try and integrate the many different facets of statistics and research will find that Abelson has done a magnificent job supported by enlightening examples and comprehensive references.

This is a great book. Everyone who uses statistics in any way should read it. Maybe everyone who READS articles that contain statistics should read it! The mathematics is minimal (very few formulas, and those are basic), but a lot of very good advice on how to use statistics sensibly (and how it is sometimes used nonsensically!).