Biased Regression: A Ten year Perspective,

Abstract

Biased estimation in regression has experienced a tremendous growth in popularity since Hoerl and Kennard's formalization of ridge regression. Along with the interest in biased regression, many research efforts over the last ten years have extended both the theory and application of these methodologies. So too, criticisms have arisen which focus on the incompleteness of the theoretical results and on exaggerated claims about the merits of biased estimators. Rather than attempting to arbitrate these opposing views, this article discusses biased estimation with special emphasis on its ultimate justification: application to real problems. Advantages and disadvantages of three biased estimators (principal component, latent root, and ridge regression estimators) are discussed and illustrated through a comprehensive analysis of a data set on automobile emissions. From the discussion and analysis it is hoped that a more balanced perspective on the application of biased estimation will be fostered. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 1980
Accession Number
ADA091551

Entities

People

  • Richard F. Gunst

Organizations

  • Southern Methodist University

Tags

Communities of Interest

  • Engineered Resilient Systems

DTIC Thesaurus Topics

  • Air Force
  • Anomaly Detection
  • Barometric Pressure
  • Celestial Brightness
  • Change Detection
  • Chemical Engineering
  • Coefficients
  • Data Sets
  • Databases
  • Emission
  • Estimators
  • Information Science
  • Regression Analysis
  • Simulations
  • Statistics
  • United States
  • United States Government

Readers

  • Educational Psychology
  • Regression Analysis.