Identifying Rank-Influential Groups of Observations in Linear Regression Modeling (Classic Reprint) - Softcover

Peter J. Kempthorne

 
9781332264612: Identifying Rank-Influential Groups of Observations in Linear Regression Modeling (Classic Reprint)

Synopsis

This book explores the identification of groups of observations that exert a strong influence on the results of linear regression modeling. The author presents mathematical characterizations of these groups and proposes methods using factor-analytic rotation algorithms and projection pursuit techniques. One method focuses on finding directions in which many cases are extreme inliers, while the other examines directions in which most cases are outlying. Both methods avoid a combinatorial search through possible subsets of cases, making them computationally feasible for data sets of various sizes. The author provides examples to illustrate how the methods perform, and the book concludes with a discussion of related diagnostic methods. The insights provided by this book are significant for researchers and practitioners who use linear regression modeling to analyze data, as they highlight the importance of considering the impact of influential groups of observations on the reliability and validity of their results.

"synopsis" may belong to another edition of this title.

Other Popular Editions of the Same Title