Open Conference Systems, 50th Scientific meeting of the Italian Statistical Society

Font Size: 
Power Priors for Bayesian Analysis of Graphical Models of Conditional Independence in Three Way Contingency Tables
AIKATERINI MANTZOUNI

Last modified: 2018-05-18

Abstract


In this paper, we illustrate a comprehensive Bayesian analysis of graphical models of conditional independence, involving suitable choices of prior parameters, estimation, model determination, as well as the allied computational issues for three-way contingency tables. Each conditional independence model corresponds to a particular factorization of the cell probabilities and a conjugate analysis based on Dirichlet prior can be performed. Unit information interpretation priors are used as a yardstick in order to identify and interpret the effect of any other prior distribution used. The posterior distributions of the graphical models parameters, are obtained using simple Markov chain Monte Carlo (MCMC) schemes.  A real data application will be presented during the talk

References


  • Bartlett, M. S.: A comment on Lindley's statistical paradox. Biometrika 44, pp. 533-534. (1957)
  • Chen, M.H., Ibrahim, J.G. and  Shao, Q. M. Power prior distributions for generalized linear models. Journal of Statistical Planning and Inference 84, pp. 121-137. (2000)
  • Darroch, J.N., Lauritzen, S.L., and Speed, T.P.: Markov Fields and Log-Linear Interaction Models for Contingency Tables.Annals of  Statistics 8, pp. 522-539. (1980)
  • Dawid A.P. and   Lauritzen S.L.: Hyper-Markov laws in the statistical analysis of decomposable graphical models.Annals of  Statistics 21, pp. 1272-1317.(1993)
  • Dawid, A.P., and Lauritzen, S.L.: Compatible prior distributions. In Bayesian Methods with Applications to Science Policy and Official Statistics. The sixth world meeting of the International Society for Bayesian Analysis (ed. E.I.George), pp. 109-118.(2000) http://www.stat.cmu.edu/ISBA/index.html|.
  • Dellaportas, P. and Forster, J.J.: Markov Chain Monte Carlo Model determination for hierarchical and graphical log-linear Models. Biometrika 86, pp. 615-634.(1999)
  • Dellaportas, P., Forster, J.J. and Ntoufras I.: On Bayesian model and variable selection using MCMC. Statistics and Computing 12, pp. 27-36.(2002)
  • Ibrahim J.G. and Chen M. H.: Power PriorDistributions for Regression Models. Statistical Science 15, pp. 46-60.(2000)
  • Lindley, D. V.: A statistical paradox.Biometrika, 44, 187-192. (1957)
  • Perks, W.: Some observations on inverse probability including a new indifference rule. Journal of the Institute of Actuaries 73, pp. 285-334. (1947)
  • Roverato A. and Consonni G.: Compatible PriorDistributions for DAG models. Journal of the Royal Statistical Society B 66, pp. 47-61. (2004)
  • Steck, H. and   Jaakkola, T.: On the Dirichlet Prior and Bayesian Regularization. NIPS, pp. 697-704. (2002)
  • Steck, H.:  Learning the Bayesian Network Structure: Dirichlet Prior vs Data. Proceedings of the conference on Uncertainty in Artificial Intelligence, pp. 511-518. (2008)
  • Hamburger, C.: Quasimonotonicity, regularity, and duality for nonlinear systems of partial differential equations. Ann. Mat. Pura. Appl. 169, pp. 321-354. (1995)
  • Ueno, M.: Learning Likelihood-Equivalence BayesianNetworks Using an Empirical Bayesian Approach. Behaviormetrika 35 pp. 115-135.(2008)

Full Text: RAR