Skip to content

Research at St Andrews

Exploring dependence between categorical variables: benefits and limitations of using variable selection within Bayesian clustering in relation to log-linear modelling with interaction terms

Research output: Contribution to journalArticle


Michail Papathomas, Sylvia Richardson

School/Research organisations


This manuscript is concerned with relating two approaches that can be used to explore complex dependence structures between categorical variables, namely Bayesian partitioning of the covariate space incorporating a variable selection procedure that highlights the covariates that drive the clustering, and log-linear modelling with interaction terms. We derive theoretical results on this relation and discuss if they can be employed to assist log-linear model determination, demonstrating advantages and limitations with simulated and real data sets. The main advantage concerns sparse contingency tables. Inferences from clustering can potentially reduce the number of covariates considered and, subsequently, the number of competing log-linear models, making the exploration of the model space feasible. Variable selection within clustering can inform on marginal independence in general, thus allowing for a more efficient exploration of the log-linear model space. However, we show that the clustering structure is not informative on the existence of interactions in a consistent manner. This work is of interest to those who utilize log-linear models, as well as practitioners such as epidemiologists that use clustering models to reduce the dimensionality in the data and to reveal interesting patterns on how covariates combine.


Original languageEnglish
Pages (from-to)47-63
JournalJournal of Statistical Planning and Inference
Early online date15 Jan 2016
Publication statusPublished - Jun 2016

    Research areas

  • Bayesian model selection, Sparse contingency tables, Graphical models

Discover related content
Find related publications, people, projects and more using interactive charts.

View graph of relations

Related by author

  1. On the correspondence from Bayesian log-linear modelling to logistic regression modelling with g-priors

    Papathomas, M., Mar 2018, In : TEST. 27, 1, p. 197-220 24 p.

    Research output: Contribution to journalArticle

  2. Primary education in Vietnam and pupil online engagement

    Nguyen, Q., Naguib, R., Das, A., Papathomas, M., Vallar, E., Wickramasinghe, N., Santos, G. N., Galvez, M. C. & Nguyen, V., 8 Jan 2018, In : International Journal of Educational Management. 32, 1, p. 71-83

    Research output: Contribution to journalArticle

  3. Parameter redundancy and the existence of the maximum likelihood estimates in log-linear models

    Sharifi Far, S., Papathomas, M. & King, R., 2018, (Submitted) In : Mathematics arXiv.

    Research output: Contribution to journalArticle

Related by journal

  1. Design and analysis of experiments testing for biodiversity effects in ecology

    Bailey, R. A. & Reiss, J., Jan 2014, In : Journal of Statistical Planning and Inference. 144, p. 69-80 12 p.

    Research output: Contribution to journalArticle

  2. A note on robustness of D-optimal block designs for two-colour microarray experiments

    Bailey, R. A., Schiffl, K. & Hilgers, R-D., 2013, In : Journal of Statistical Planning and Inference. 143, p. 1195-1202 18 p.

    Research output: Contribution to journalArticle

  3. Steady-state Markov chain models for certain q-confluent hypergeometric distributions

    Kemp, A. W., 1 Nov 2005, In : Journal of Statistical Planning and Inference. 135, 1, p. 107-120 14 p.

    Research output: Contribution to journalArticle

ID: 240080580