Model based biclustering

The problem of biclustering has arisen in diverse research areas, such as simultaneous grouping of genes (or proteins) and experimental conditions in functional genomics, co-clustering of documents and words in text mining, co-characterisation of users and movies in recommender systems, and analysis of shopping attitudes of customers in market research. Various procedures have been developed to cope with this problem. Here, we address this problem by using a family of semiparametric plaid models under the assumption that the variances of observations are power functions of their means. The performance of our proposal has been assessed by both simulation studies and real data applications.