Logic of Simpson paradox

Jacek Malinowski

DOI: http://dx.doi.org/10.12775/LLP.2005.013

Abstract


The main aim of this paper is to elucidate, from a logical point of view, the phenomenon of Simpson reversal — the paradox of a statistical reasoning. We define a binary relation of supporting in the following way: a sentence A supports a sentence B if and only if the probability of B is higher when A is true, than when A is false. It appears that a statistical argument occurring in Simpson paradox cannot be formalized by means of a binary relation. We generalize the relation of support introducing the third parameter. Then we argue that it properly mirrors main features of the statistical argument occurring in Simpson paradox.

Keywords


logical entailment; statistical inference; Bayesian inference; Simpson paradox

Full Text:

PDF

References


Carnap, R. [1951], Logical foundations of probability, Routledge and Kegan Paul, London.

Howson, C., and P. Urbach [1989], Scientific Reasoning: the Bayesian Approach, Open Court Publishing Co., La Salle, Illinois.

Kuipers, T. [2000], From Instrumentalism to Constructive Realism: On Some Relations Between Confirmation, Empirical Progress, and Truth Approximation, Synthese Library 287, Kluwer Academic Press, Dordrecht.

Makinson, D. [2005], Bridges from Classical to Nonmonotonic Logic, Texts in Computing, Kings College, London.

Malinowski, J. [2005], “Bayesian propositional logic”, to appear.

Pearl, J. [2000], Causality: Models reasoning, and inference, Cambridge University Press, Cambridge.

Popper, K. [1968], The Logic of Scientific Discovery, 3rd. (revised) edition, Hutchinson, London.

Simpson, E.H. [1951], “The Interpretation of Interaction in Contingency Table”, Journal of Royal Statistical Society, Series B, vol. 13, 238–241.








ISSN: 1425-3305 (print version)

ISSN: 2300-9802 (electronic version)

Partnerzy platformy czasopism