Model-Based Block Clustering with EM Algorithm

• Activity Number: 463
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Prabhani Kuruppumullage Don*+ and Bruce G. Lindsay and Francesca Chiaromonte
• Companies: Penn State and Penn State and Penn State
• Keywords: Block clustering ; EM algorithm ; Composite likelihood ; Gradient function

Abstract

With the increase interest in two-way clustering methods, there have been various attempts to develop mixture-based approaches. Computational challenges in these settings have led scientists to propose various approximations to the mixture likelihood. In this work, we propose an approach based on a composite likelihood approximation and a nested EM algorithm to maximize the likelihood. Further, We devise the gradient function for our setting and discuss how one can evaluate the suitability of the number of components by using an EM algorithm on the mixture gradient function. Finally, we combine the EM on gradient function with a bootstrap-based composite likelihood ratio test to help in model selection in this setting.