Title: Analyzing large matrices of ordinal data
Authors: Julien Jacques - Universite de Lyon (France)
Margot Selosse - Universite de Lyon (France) [presenting]
Christophe Biernacki - Inria (France)
Abstract: A co-clustering strategy for analyzing large matrix of ordinal data is presented. For this, a model-based co-clustering algorithm for ordinal data is proposed. This algorithm relies on the latent block model embedding a probability distribution specific to ordinal data (the so-called BOS or Binary Ordinal Search distribution). Model inference relies on a stochastic EM algorithm coupled with a Gibbs sampler, and the ICL-BIC criterion is used for selecting the number of co-clusters (or blocks). The main advantage of this ordinal dedicated co-clustering model is its parsimony, the interpretability of the co-cluster parameters (mode, precision) and the possibility to take into account missing data. The usefulness of the method is illustrated by analyzing a psychological survey on women affected by a breast tumor.