Statistical Data Analysis Handbook

By Francis J. Wall

No Synopsis Available.

• Format: Hardcover
• Language: English
• ISBN: 0070679312
• ISBN13: 9780070679313
• Release Date: January 1985
• Publisher: McGraw-Hill Companies
• Length: 285 Pages
• Weight: 1.75 lbs.

Customer Reviews

Revealing Facts from Statistical Data

The author's aim in writing the book is to show the naïve data analyst how to use statistical methods by focusing his attention and experience on a wide variety of practical problems. However, do not be misled by the title of this book. The book is a collection of real data and some statistical treatment on them with varying degree of detail. The book is not sufficiently comprehensive to be used as a reference book for the practicing statistician. As it is, the book does not serve as a handbook in data analysis. The book contains five parts. The first three parts present the topics in the context of analyzing detailed practical problems. Part I: Simple comparisons - It deals with the t test for independent and paired data, and ANOVA. For these parametric techniques the related nonparametric procedures are also discussed. Analysis sequence flowcharts are also given. Part II: Linear relationships - It gives an introduction to simple, polynomial and multivariate linear regression, and in Part II: Comparison of relationships - Comparisons of two or more than two linear relationships are discussed. Part IV: Supporting topics ? It contains a survey, reviewing introductory statistics for those who have not already had exposure to statistical theory. Part V: Tables ? containing 20 tables mostly arranged in terms of the upper tail probabilities (P-values). The first four parts are divided into 20 chapters paged within each chapter. Some chapters are very brief. I have preferred Chapter 14, which covers transformations, to be omitted since it is too brief. Thus all the references to this chapter (e. g., pp. 2.13-2.18) throughout the book should be omitted. Certain other important topics have not been adequately covered or discussed. Particular omissions which might be seen as important include: analysis of residuals, remedies for violations of regression assumption, the treatment of outliers, autoregressive error structures for time series data (example 10.1) and the Durbin-Watson test, and the effect of sample size on the meaning of the P value which is an issue since some of the data sets are small. These topics are closely intertwined, however, and the model building chapters are incomplete without the residual analysis. In addition, there are a few mistakes or ambiguities. These could cause real problems for readers approaching the subject for the first time. These topics include the following: (a) The computation of Lilliefors statistic for data sets having some observations with equal values (e.g. Table 15.6) is incorrect. This error is caused by miscalculating the sample probabilities. Although this correction has no impact on the results, I found it rather tedious to compute this statistic and prefer to use the readily available acceptance graphs for this test. (b) The test for equality of variances assumes normality which must be tested first. All the given flowcharts must be corrected to reflect this fact. (