HOW TO LOOK AT DATA: A REVIEW OF JOHN W. TUKEY'S EXPLORATORY DATA ANALYSIS1

Here is a book about data analysis that should be fascinating to many readers of the Journal of the Experimental Analysis of Behavior. It is quite different from any standard textbook of statistics: It does not deal with testing hypotheses and establishing confidence intervals. Instead, it provides techniques and advice about how to explore data. The approach is quite compatible with the practices of many readers of JEAB. The book describes many specific methods and a general approach. The approach of exploratory data analysis is described as being detective in character. It is a search for clues. Some of the clues may be misleading, but some will lead to discoveries. After the appearances are described, other techniques can be employed for purposes of confirmation, but this book deals only with the exploratory phase. It is necessary to discover facts before they can be confirmed. The author, a distinguished statistician, clearly expresses his values. He favors simplicity because simple statements are clear. He particularly features clear visual displays of quantitative facts. He likes precision because a more precise statement contains more information than a less exact statement. For example, it is far better to be able to say some response measure is a linear function of a particular stimulus variable than to say it increases with the stimulus variable. He likes flexibility of approach because it is seldom clear exactly which methods will best achieve the goals of the data analyst, and sometimes different analyses of the same data reveal different aspects of it. He favors depth of analysis. It is always good to look at the residuals;