Previous |  Up |  Next


contingency tables; hierarchical models; partial maximization algorithms
In this paper we analyze the asymptotic behavior of the IPF algorithm for the problem of finding a 2x2x2 contingency table whose pair marginals are all equal to a specified 2x2 table, depending on a parameter. When this parameter lies below a certain threshold the marginal problem has no solution. We show that in this case the IPF has a “period three limit cycle” attracting all positive initial tables, and a bifurcation occur when the parameter crosses the threshold.
[1] Csiszár I.: I-divergence geometry of probability distributions and minimization problems. Ann. Probab. 3 (1975), 146–158 DOI 10.1214/aop/1176996454 | MR 0365798
[2] Deming W. E., Stephan F. F.: On a least square adjustment of a sampled frequency table when the expected marginal totals are known. Ann. Math. Statist. 11 (1940), 427–444 DOI 10.1214/aoms/1177731829 | MR 0003527
[3] Haberman S. J.: The analysis of frequency data. The University of Chicago Press, Chicago 1974 MR 0408098 | Zbl 0325.62017
[4] Jensen S. T., Johansen, S., Lauritzen S. L.: Globally convergent algorithms for maximizing a likelihood function. Biometrika 78 (1991), 867–877 MR 1147024 | Zbl 0752.62031
[5] Jiroušek R.: Solution of the marginal problem and decomposable distributions. Kybernetika 27 (1991), 403–412 MR 1132602 | Zbl 0752.60009
[6] Lauritzen S. L.: Graphical Models. Clarendon Press, Oxford 1996 MR 1419991 | Zbl 1055.62126
Partner of
EuDML logo