Abstract
We propose an alternative resolution of Simpson's paradox in multiple classification experiments, using a different maximum likelihood estimator. In the center of our analysis is a formal representation of free choice and randomization that is based on the notion of incompatible measurements.
We first introduce a representation of incompatible measurements as a collection of sets of outcomes. This leads to a natural generalization of Kolmogoroff's axioms of probability. We then discuss the existence and uniqueness of the maximum likelihood estimator for a probability weight on such a generalized sample space, given absolute frequency data.
As a first example, we discuss an estimation problem with censured data that classically admits only biased ad hoc estimators.
Next, we derive an explicit solution of the maximum likelihood estimation problem for a large class of experiments that arise from various kids of compositions of sample spaces. We identify the (categorical) direct sum of sample spaces as a representation of “free choice,” and the (categorical) direct product as a representation of “randomization.”
Finally, we apply the foregoing discussion to the case of multiple classification experiments in order to show that there is no Simpson's paradox if the difference between free choice and randomization is recognized in the structure of the experiment.
A comparison between our new estimator and the “usual” calculation can be summarized as follows: Pooling the data over one classification factor in the “usual” way in fact destroys or ignores the information contained in it, whereas our proposed maximum likelihood estimator is a proper marginal over this factor that “averages out” the information contained in it. The estimators agree with each other in the case of proportional sample sizes.
Similar content being viewed by others
References
P. J. Bickel, E. A. Hammel, and J. W. O'Connell, “Sex bias in graduate admissions: Data from Berkeley,”Science 187, 398–404 (1975).
C. R. Blyth, “On Simpson's paradox and the sure-thing principle,”J. Am. Stat. Assoc. 67, 364–366 (1972).
C. R. Blyth, “Simpson's paradox and mutually favorable events,”J. Am. Stat. Assoc. 68, 746 (1973).
N. Cartwright,How the Laws of Physics Lie (Clarendon, New York, 1983).
M. R. Cohen and E. Nagel,An Introduction to Logic and Scientific Method (Harcourt, Brace, and Co, New York, 1934).
J. C. Dacey, Jr., “Orthomodular spaces,” Dissertation, University of Massachusetts, Amherst, 1968.
D. J. Foulis and C. H. Randall, “Operational statistics. I. Basic concepts,”J. Math. Phys. 13, 1667–1675 (1972).
D. J. Foulis and C. H. Randall, “Empirical logic and quantum mechanics,”Synthese 29, 81–111 (1974).
D. J. Foulis and C. H. Randall, “Empirical logic and tensor products,” inInterpretations and Foundations of Quantum Theory, H. Neumann, ed. (Bibliographisches Institut, Mannheim, 1981), pp. 9–20.
D. Freedman, R. Pisani, and R. Purves,Statistics (Norton, New York, 1978).
M. Gaudard, “The prior in Bayesian inference: a nonclassical approach,” Dissertation, University of Massachusetts, Amherst, 1977.
M. Gaudard, “Convergence of posterior probabilities in the Bayesian inference strategy,”Found. Phys. 15, 49–62 (1985).
E. R. Gerelle, R. J. Greechie, and F. R. Miller, “Weights on spaces,” inPhysical Reality and Mathematical Description, E. P. Enz and J. Mehra, eds. (Reidel, Dordrecht, 1974), pp. 169–192.
R. J. Greechie and S. P. Gudder, “Quantum logics,” inContemporary Research in the Foundations and Philosophy of Quantum Theory, C. A. Hooker, ed. (Reidel, Dordrecht, 1973), pp. 143–173.
S. P. Gudder, “Logical cover spaces,”Ann. Inst. Henri Poincaré 45, 327–337 (1986).
S. P. Gudder, M. P. Kläy, and G. T. Rüttimann, “States on hypergraphs,”Dem. Math. 19, 503–526 (1986).
S. P. Gudder, G. T. Rüttimann, and R. J. Greechie, “Measurement, Hilbert space and quantum logic,”J. Math. Phys. 23, 2381–2386 (1982).
M. P. Kläy, “Stochastic models on empirical systems, empirical logics and quantum logics, and states on hypergraphs,” Dissertation, Universität Bern (1985).
M. P. Kläy, “Quantum logic properties of hypergraphs,”Found. Phys. 17, 1019–1036 (1987).
M. P. Kläy, C. H. Randall, and D. J. Foulis, “Tensor products and probability weights,”Int. J. Theor. Phys. 26, 199–219 (1987).
T. R. Knapp, “Instances of Simpson's paradox,”Coll. Math. J. 16, 209–211 (1985).
A. Kolmogoroff,Grundbegriffe der Wahrscheinlichkeitsrechnung (Springer, Berlin, 1933).
J. von Neumann,Mathematische Grundlagen der Quantenmechanik (Springer, Berlin, 1932).
C. H. Randall and D. J. Foulis, “An approach to empirical logic,”Am. Math. Mon. 77, 363–373 (1970).
C. H. Randall and D. J. Foulis, “Operational statistics. II. Manuals of operations and their logics,”J. Math. Phys. 14, 1472–1480 (1973).
C. H. Randall and D. J. Foulis, “A mathematical setting for inductive reasoning,” inFoundations of Probability Theory, Statistical Inference, and Statistical Theories of Science, Vol. III, W. L. Harper and C. A. Hooker, eds. (Reidel, Dordrecht, 1976), pp. 169–205.
C. H. Randall and D. J. Foulis, “Operational statistics and tensor products,” inInterpretations and Foundations of Quantum Theory, H. Neumann, ed. (Bibliographisches Institut, Mannheim, 1981), pp. 21–28.
C. H. Randall and D. J. Foulis, “A mathematical language for quantum physics,” inTransactions, 25e Cours de perfectionnement de l'Association Vaudoise des Chercheurs en Physique: Les fondements de la Mécanique Quantique, C. Gruber, C. Piron, T. Minhtôm, and R. Weill, eds. (Montana, Switzerland, 1983), pp. 193–226.
G. T. Rüttimann, “Detectable properties and spectral quantum logics,” inInterpretations and Foundations of Quantum Theory, H. Neumann, ed. (Bibliographisches Institut, Mannheim, 1981), pp. 35–47.
E. H. Simpson, “The interpretation of interaction in contingency tables,”J. R. Stat. Soc., Ser. B 13, 238–241 (1951).
C. H. Wagner, “Simpson's paradox in real life,”Am. Stat. 36, 46–48 (1982).
G. U. Yule, “Notes on the theory of associations of attributes in statistics,”Biometrika 2, 121–134 (1903).
Author information
Authors and Affiliations
Additional information
Work supported by the Swiss National Science Foundation.
Rights and permissions
About this article
Cite this article
Kläy, M.P., Foulis, D.J. Maximum likelihood estimation on generalized sample spaces: An alternative resolution of Simpson's paradox. Found Phys 20, 777–799 (1990). https://doi.org/10.1007/BF01889691
Received:
Issue Date:
DOI: https://doi.org/10.1007/BF01889691