A Modern Approach to Regression with R

Abstract

A Modern Approach to Regression with R is aimed, according to its preface, at first-year graduate students and senior undergraduates. This book fills an important niche in the regression textbook by providing a data-centered approach strong on graphics. Modern Approach is very much in the spirit of Cook and Weisberg (1999) and Weisberg (2005).I have taught an undergraduate applied regression course for several years, and found few textbooks that are appropriate for modern students. To reveal my prejudices up front, what I want in an applied regression text are data, graphics, and computation. Data, because an applied regression course should give students an approach for understanding the world as it is revealed through data; the regression model is an end to a means but should not be treated as the end itself. Graphics, because not only are they important to modern regression, but they can help students develop a conceptual understanding of regression models. Computation, because data analysis via graphical and numerical regression is not possible without statistical software. The choice of software not only determines the sort of diagnostics and visualizations available to one's students, but also influences how students understand regression itself.Sheather defines regression as "a method for investigating the functional relationship among variables". This definition puts the emphasis in exactly the right place and signals that the book is teaching a way of examining data, and is not a study of the linear model itself. An emphasis on real-world problems is carried throughout the book. A great strength of this book is the numerous case studies, which make for interesting reading and always include a clearly stated goal. Four primary case studies are introduced in great detail in the first chapter, and later chapters expand on these studies and introduce new problems, too. The examples are diverse: NFL kickers, wine critics, newspaper circulation, restaurant menus.Modern Approach differs from some other books in that it provides a complete examination of simple regression (including weighted least squares) before beginning multiple regression. Chapters 2-4 cover all aspects of simple regression, and these chapters mirror the development of the multiple regression in the rest of the book. When teaching undergraduates, as I do, this means a substantial part of the ten-week term is spent on simple linear regression. This

Estimation of Forest Structural Diversity Using the Spectral and Textural Information Derived from SPOT-5 Satellite Images

“…In order to correct nonlinearity and non-constant variance, we used Box-Cox transformations of the response variables (forest stand variables). For the determination of the potential subset of independent variables, there are two distinctly different approaches, namely, all possible subsets and stepwise methods [54]. In the present study, we used all possible subsets.…”

Successful Transition to Adulthood: Do We Really Need to Teach Shopping, Sex and Social Skills?

“…To mitigate the bias of small sample size and increase reliability of our generalized results, we conducted a re-sampling bootstrap method of analysis. This estimation method allows statistical analysis of small sample sizes, as it repeats 1000 times our study parameters using random samples of our original participants (Chernick, 2008;Singh & Xie, 2008;Scholz, 2007;Sheather, 2009). These bootstrapped results are recorded throughout the paper.…”

Stable arrangements of mobile sensors for sampling physical fields

“…where € B = Φ T WΦ and w i 's form the main diagonal of the weight matrix W [30]. Typically, weights can be chosen in proportion to the separation between points, for example half the nearest neighbor distance of a sampling point [31].…”