How to cite this WIREs title:
WIREs Comp Stat

Simpson's paradox

Focus Article

Aylin Alin

Published Online: Mar 19 2010

DOI: 10.1002/wics.72

Full article on Wiley Online Library: HTML PDF

Can't access this content? Tell your librarian.

Abstract Simpson's paradox occurs when an observed association between two variables is reversed after considering the third variable. Having two different conclusions makes this phenomenon paradoxical. In this article, it will be shown that the source of this paradox is the interactions between the third variable and any of the other two variables. There will also be a section about which association should be trusted. Copyright © 2010 John Wiley & Sons, Inc. This article is categorized under: Statistical Models > Classification Models

References

Pearl, J. Why there is no statistical test for confounding, why many think there is, and why they are almost right. UCLA Cognitive Systems Laboratory Technical Report R‐256. 1998.
Schield, M. Simpson`s paradox and Cornfield`s conditions. In: Proceedings of the ASA‐JSM Section of Statistical Education, ASA. 1999, 106–111.
Simpson, EH. The interpretation of interaction in contingency tables. J Roy Stat Soc B 1951, 13: 238–241.
Fienberg, SE. The Analysis of Cross‐Classified Categorical Data. Massachusetts: The MIT Press; 1994.
Agresti, A. Categorical Data Analysis. New York: John Wiley %26 Sons; 1990.
Yule, GU. Notes on the theory of association of attributes in statistics. Biometrika 1903, 2: 121–134.
Blyth, CR. On Simpson`s paradox and the sure‐thing principle. J Am Stat Assoc 1972, 67: 364–366.
Bishop, MM, Fienberg, SE, Holland, PW. Discrete Multivariate Analysis: Theory and Practice. Massachusetts: The MIT Press; 1975.
Reintjes, R, De Boer, A, Van Pelt, W, De Groot, JM. Simpson`s paradox: an example from hospital epidemiology. Epidemiology 2000, 11: 81–83.
Westbrooke, I. Simpson`s paradox: an example in a New Zealand survey of jury composition. Chance 1998, 11: 40–42.
Heydtmann, M. The nature of truth: Simpson`s paradox and the limits of statistical data. Qxf J Med 2002, 95: 247–249.
Cornfield, J, Haenszel, W, Hammond, E, Lilienfeld, A, Shimkin, M, et al. Smoking and lung cancer: recent evidence and a discussion of some questions. J Natl Cancer Inst 1959, 22: 173–203.
Lindley, DV, Novick, MR. The role of exchangeability in inference. Ann Stat 1981, 9: 45–58.
Pearl, J. Simpson`s paradox: a anatomy. UCLA Cognitive Systems Laboratory Technical Report R‐264. 1999.
Pearl, J. Causal diagrams for empirical research. Biometrika 1995, 82: 669–710.
Otte, R. Probabilistic causality and Simpson`s paradox. Phil Sci 1985, 52: 110–125.

Resources

Further Readings

Agresti A. Fallacies, statistical. In: Kotz S, Johnson NL, Editors, Encyclopedia of Statistical Science. New York: Wiley; 1983, 3: 24-28.
Samuels ML. Simpson’s paradox and related phenomena. Journal of the American Statistical Association 1993, 88(421): 81-88.
Pavlides MG, Perlman MD. How likely is Simpson’s paradox? The American Statistician 2009, 63(3): 226-233.
Chen A, Bengtsson T, Ho TK. A regression paradox for linear models: sufficient conditions and relation to Simpson’s paradox. The American Statistician 2009, 63(3):
218-225.

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

Simpson's paradox

Browse by Topic

Access to this WIREs title is by subscription only.

The latest WIREs articles in your inbox