Shrinkage Estimation Of A Linear Regression

Shrinkage Estimation of a Linear Regression

This book deals with shrinkage regression estimators obtained by shrinking the ordinary least squares (OLS) estimator towards the origin. The author's main concern is to compare the sampling properties of a family of Stein-rule estimators with those of a family of minimum mean squared error estimators. In this book, the author deals with shrinkage regression estimators obtained by shrinking the ordinary least squares (OLS) estimator towards the origin. In particular, he deals with a family of Stein-rule (SR) estimators and a family of minimum mean squared error (MMSE) estimators.

Shrinkage Estimation

This book provides a coherent framework for understanding shrinkage estimation in statistics. The term refers to modifying a classical estimator by moving it closer to a target which could be known a priori or arise from a model. The goal is to construct estimators with improved statistical properties. The book focuses primarily on point and loss estimation of the mean vector of multivariate normal and spherically symmetric distributions. Chapter 1 reviews the statistical and decision theoretic terminology and results that will be used throughout the book. Chapter 2 is concerned with estimating the mean vector of a multivariate normal distribution under quadratic loss from a frequentist perspective. In Chapter 3 the authors take a Bayesian view of shrinkage estimation in the normal setting. Chapter 4 introduces the general classes of spherically and elliptically symmetric distributions. Point and loss estimation for these broad classes are studied in subsequent chapters. In particular, Chapter 5 extends many of the results from Chapters 2 and 3 to spherically and elliptically symmetric distributions. Chapter 6 considers the general linear model with spherically symmetric error distributions when a residual vector is available. Chapter 7 then considers the problem of estimating a location vector which is constrained to lie in a convex set. Much of the chapter is devoted to one of two types of constraint sets, balls and polyhedral cones. In Chapter 8 the authors focus on loss estimation and data-dependent evidence reports. Appendices cover a number of technical topics including weakly differentiable functions; examples where Stein’s identity doesn’t hold; Stein’s lemma and Stokes’ theorem for smooth boundaries; harmonic, superharmonic and subharmonic functions; and modified Bessel functions.

Generalized Shrinkage Estimation for Multiple Linear Regression