bikm1:Co-Clustering Adjusted Rand Index and Bikm1 Procedure for
Contingency and Binary Data-Sets
Co-clustering of the rows and columns of a contingency or binary matrix, or double binary matrices and model selection
for the number of row and column clusters. Three models are
considered: the Poisson latent block model for contingency
matrix, the binary latent block model for binary matrix and a
new model we develop: the multiple latent block model for
double binary matrices. A new procedure named bikm1 is
implemented to investigate more efficiently the grid of numbers
of clusters. Then, the studied model selection criteria are the
integrated completed likelihood (ICL) and the Bayesian
integrated likelihood (BIC). Finally, the co-clustering
adjusted Rand index (CARI) to measure agreement between
co-clustering partitions is implemented. Robert Valerie,
Vasseur Yann, Brault Vincent (2021)
<doi:10.1007/s00357-020-09379-w>.
