The monograph belongs to the series Texts in Statistical Science and presents the sixth upgraded edition of the popular manual. It was first issued in 1984, and from that time won recognition as one of the best textbooks on the applied statistical modeling and analysis. The book is organized in two parts and eighteen chapters: the first part considers “Preparation for Analysis” (Chapters 1–6), and the second part is called “Regression Analysis” (Chapters 7–18) although it covers many other related methods as well.
Chapter 1 defines multivariate analysis, in its exploratory and confirmatory aspects, gives examples, and draws the book structure. Chapter 2 characterizes data of different types, such as nominal, ordinal, interval, ratio, categorical, continuous, discrete, explanatory, and dependent. Chapter 3 deals with preparing data for analysis, briefing on statistical software of R, SAS, SPSS, and Stata packages, describes techniques of data entry, problems with missing data and outliers. Added in this edition new Chapter 4 discusses various methods of data visualization with specific tools for different types of bivariate and multivariate variables, and gives multiple illustrations in graphs and figures. Chapter 5 presents possible ways for data screening and transformation, needed for normality of distribution used in many hypotheses testing. Chapter 6 describes selection of appropriate methods of multivariate analysis due to the data types and purposes of research.
Chapter 7 presents simple linear regression, its residual errors, confidence intervals, parameter hypotheses testing, correlation, fixed predictor variable, bivariate regression as the conditional distribution. Linearized models by transformed variables and computer programs are also discussed. Chapter 8 considers multiple linear regression with its various features and characteristics of fit quality, tests on parameters, and extensions to polynomial models. Chapter 9 describes methods of variable selection in regression, including Akaike and Bayesian information criteria, F-criterion in stepwise modeling, and Lasso regression. Chapter 10 covers special regression topics, such as missing values multiple imputation, dummy or indicator variables, constraints on parameters, multicollinearity, and ridge regression. Chapter 11 focuses on discriminant analysis and classification into groups by Fisher function, adjusting the dividing points, measuring goodness of fit, cross-validation, and jackknife procedure. Chapter 12 continues with logistic regression used for classification and other purposes, describes receiver operating characteristic, or ROC curve, nominal and ordinal logistic models, Poisson regression, and generalized linear model, or GLM. Chapter 13 deals with regression for survival data, describing hazard functions for exponential and Weibull distributions, log-linear model, and Cox proportional hazard regression. Chapter 14 presents principal components analysis, its features and applications. Chapter 15 describes factor analysis in its exploratory version, with techniques of factors rotation, and interpretation of the results. Chapter 16 presents cluster analysis in its several algorithms by agglomerative and divisive methods, including K-means and hierarchical clustering. Chapter 17 is devoted to the log-linear analysis for two-way and multi-way tables, with stepwise selection, assessing specific models, and comparison with the logit models. Chapter 18 concludes consideration by the correlated outcomes regressions, when the dependent variable observations are already not independent among themselves but can be viewed as related in subgroups or by measurements in longitudinal studies. Conditional and marginal models with fixed and random effects are described, and generalized estimating equations, GEE, are discussed as well. The book is finalized by the Appendix containing references to the data sources used in the examples, bibliography of more than three hundred sources, and a detail index given in eight pages.
Most of chapters of the first part of the textbook contain such subsections as “Introduction” or “Definition,” “Discussion” or “Examples,” “Summary” and “Problems.” And almost all chapters of the second part of the textbook start with the subsections of “Chapter Outline,” “When This Technique Is Used,” “Data Example,” “Basic Concepts,” and finish with “Discussion of Computer Programs,” “What to Watch Out For,” “Summary,” and “Problems” suggesting a couple of dozen of exercises for each chapter. This structure makes the book very reader-friendly written, helping to students and researchers in various fields to understand what for a statistical tool can serve, how to apply it, and to interpret computer outputs. There is not much of mathematical and statistical derivation, neither modern statistical techniques, but plenty of examples oriented to the easy “know-how” practical implementations of the classical multivariate methods.
Minneapolis, MN