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

Font Size: 
Bayesian Estimation of Graphical Log-Linear Marginal Models
Claudia Tarantola

Last modified: 2018-08-30


Bayesian methodsfor graphical log-linear marginal models have not been developed as much as traditional frequentist approaches.The  likelihood function cannot be analyticallyexpressed in terms of  the marginal log-linear interactions, but only in terms of cell counts or probabilities. No conjugate analysis is feasible,  and MCMC methods are needed.We present a fully automatic and efficient MCMC strategy for quantitative learning based on the DAG representation of the model. While the prior is expressed in terms of the marginal log-linear interactions, the proposal is on the probability parameter space. Furthermore, in order to obtain an efficient algorithm  we use use as proposal values draws from a Gibbs sampling on the probability parameters.


1. Bartolucci, F., Scaccia, L. and Farcomeni, A. (2012). Bayesian inference through encom-passing priors and importance sampling for a class of marginal models for categorical data.Computational Statistics and Data Analysis, 56, 4067–4080.

2. Bergsma, W. P. and Rudas, T. (2002). Marginal log-linear models for categorical data. Annalsof Statistics, 30, 140-159.

3. Bergsma,W.O.,Croon,M.A.andHagenaars,J.A.(2009).MarginalModels.ForDependent,Clustered, and Longitudinal Categorical Data, Springer

4. Drton, M. and Richardson, T. S. (2008). Binary models for marginal independence. Journal of the Royal Statistical Society, Ser. B , 70, 287-309.

5. Lupparelli, M. (2006). Graphical models of marginal independence for categorical variables.Ph. D. thesis, University of Florence.

6. Lupparelli, M., Marchetti, G. M., Bersgma, W. P. (2009). Parameterization and fitting of bi-directed graph models to categorical data. Scandinavian journal of Statistics, 36, 559-576.

7. Ntzoufras, I. and Tarantola, C. (2013). Conjugate and Conditional Conjugate Bayesian Anal-ysis of Discrete Graphical Models of Marginal Independence. Computational Statistics &Data Analysis, 66, 161-177.

8. Ntzoufras, I. Tarantola, C. and Lupparelli, M. (2018). Probability Based Independence Sam-pler for Bayesian Quantitative Learning in Graphical Log-Linear Marginal Models, DEMWorking Paper Series.

9. Richardson, T. S. (2003). Markov property for acyclic directed mixed graphs. Scandinavian Journal of Statistics 30, 145-157.

10. Silva, R. and Ghahramani, Z. (2009a). The hidden life of latent variables: Bayesian learningwith mixed graph models. Journal of Machine Learning Research 10, 1187–1238

Full Text: PDF