Censoring and Stochastic Integrals

The first edition of this book was published in 1972. This second edition is not greatly different from the first. The two editions are divided into 5 chapters and a 50-page appendix entitled, “Review of Fundamental Concepts.” The five chapter titles are the same for the two editions and are 1. “Introduction to Data Analysis,” 2. “Elementary Statistical Inference,” 3. “Regression and Correlation Analysis,” 4. “The Analysis of Variance,” and 5. “Multivariate Statistical Methods.” The second edition is 76 pages longer than the first. This is primarily due to the addition of the following sections: in Chapter 1, “Data Screening,” in Chapter 2, “Other Measures of Association for Contingency Tables” and “Robust Estimators,” and in Chapter 5, “The Multivariate Analysis of Variance;” in Chapter 3 the subsection “Multivariate Missing Observations” is added. A number of new examples and problems have been added. All chapters have a set of problems now whereas Chapter 1 did not in the first edition. No answers are provided for the problems. Most of the examples and problems are from the fields of biology and medicine and unless one is fairly familiar with these subject areas, the examples and problems will not be very meaningful. In the preface to the first edition the authors indicate that the “book contains both elementary and advanced topics” and has “a wide range of material.” In the preface to the second edition they indicate that they have attempted “to broaden the’ scope of the book.” The breadth is perhaps best illustrated by the courses that the authors suggest might be taught from the book; namely, “Elementary Applied Statistical Analysis” (undergraduate), “Applied Statistical Analysis” (first-year graduate), “Applied Multivariate Analysis” (second-year graduate), and “Intensive Course in Data Analysis.” I feel that books aimed at a specific well-defined audience are generally more effective than those aimed at a wide general audience, and this book only confirms that opinion. I tried to put myself in the place of a person who has had only an “elementary course in the fundamentals of statistical inference” and college algebra-the minimum background that the authors feel is required to use the book. I think people with that background would find the book very difficult if they were on their own. In Chapter 1 the authors say, “. there are two main objectives of this book. The first is to present in a practical manner the fundamental techniques of classical statistical analysis-both univariate and multivariate.” They have done a good job of presenting the techniques of classical statistics. However, I do not feel that they have done a good job of interpreting their examples. Most of the interpretations of the analyses consist of little more than saying whether a statistical test is significant-with a P value--or not. In the analysis of variance (ANOVA) chapter the authors introduce a study (give a literature reference) in which, “The data consisted of measurements Y = the amount of nitrogen expired (in liters) in resting conditions under four dietary regimens.” They indicate that, “This example will be analyzed in many different ways throughout the chapter.” When they use the study to illustrate the one-way ANOVA, they present some data. This study and these data (or a subset) are then used many times as examples in the chapter. I think real experimental studies and data lose their effectiveness when used this way. One might better, I feel, generate data-and admit it-and do the generation using the model to be illustrated. When a single set of data is used to illustrate successively more complex studies and ends up illustrating the situation for the real experiment that produced the data, it is sometimes effective in that it can show why the real study was set up as it was. However, in the present case it is never indicated what the design was for the study referenced. There is a brief discussion of power in the appendix. However, I find it strange that there is nothing on power in the body of the text.