Research Article

Do not adjust coefficients in Shapley value regression

  1. Ulrike Grömping1,*,
  2. Sabine Landau2

Article first published online: 17 APR 2009

DOI: 10.1002/asmb.773

Issue

Applied Stochastic Models in Business and Industry

Applied Stochastic Models in Business and Industry

Volume 26, Issue 2, pages 194–202, March/April 2010

Additional Information

How to Cite

Grömping, U. and Landau, S. (2010), Do not adjust coefficients in Shapley value regression. Applied Stochastic Models in Business and Industry, 26: 194–202. doi: 10.1002/asmb.773

Author Information

  1. 1

    Beuth Hochschule für Technik Berlin, Berlin, Germany

  2. 2

    Institute of Psychiatry, King's College London, U.K.

*Beuth Hochschule für Technik Berlin, Berlin, Germany

Publication History

  1. Issue published online: 12 APR 2010
  2. Article first published online: 17 APR 2009
  3. Manuscript Revised: 12 MAR 2009
  4. Manuscript Accepted: 12 MAR 2009
  5. Manuscript Received: 5 AUG 2008

SEARCH

SEARCH BY CITATION

ARTICLE TOOLS

Get PDF (115K)

Keywords:

  • multiple regression;
  • Shapley value regression;
  • multicollinearity;
  • Lord's paradox;
  • suppression

Abstract

Shapley value regression consists of assessing relative importance and accordingly adjusting regression coefficients. It is argued that adjustment of coefficients is unnecessary and even misleading for practically relevant situations. Examples are given, and an alternative procedure is proposed for situations for which the coefficients are requested to have a certain sign. Copyright © 2009 John Wiley & Sons, Ltd.

Get PDF (115K)