Model Choice in Nonnested Families [Paperback]

This book discusses the problem of model choice when the statistical models are separate, also called nonnested. Chapter 1 provides an introduction, motivating examples and a general overview of the problem. Chapter 2 presents the classical or frequentist approach to the problem as well as several alternative procedures and their properties. Chapter 3 explores the Bayesian approach, the limitations of the classical Bayes factors and the proposed alternative Bayes factors to overcome these limitations. It also discusses a significance Bayesian procedure. Lastly, Chapter 4 examines the pure likelihood approach. Various real-data examples and computer simulations are provided throughout the text.

Preliminaries.- Frequentist Methods.- Bayesian Methods.- Support and Simulation Methods.- Maximum Likelihood Estimation.- Index.The authors are recognized experts teaching statistics in Brazil universities, and in the book & they present various methods of choosing between competing families of regression models, for instance, exponential versus lognormal models. & The monograph is interesting, innovative, and can serve in search for adequate models in applied statistical analysis. (Stan Lipovetsky, Technometrics, Vol. 59 (4), November, 2017)

Basilio de Bragan?a Pereira is a Professor of Biostatistics and of Applied Statistics at the Federal University of Rio de Janeiro in Brazil.

Carlos Alberto de Bragan?a Pereira is a Professor of Statistics at the University of Sao Paulo in Brazil.

This book discusses the problem of model choice when the statistical models are separate, also called nonnested. Chapter 1 provides an introduction, motivating examples and a general overview of the problem. Chapter 2 presents the classical or frequentist approach to the problem as well as several alternative procedures and their properties. Chapter 3 explores the Bayesian approach, the limitations of the classicló6