B0418
Title: Robust clusterwise regression analysis for the censored data
Authors: Mehrdad Naderi - Northumbria University (United Kingdom) [presenting]
Elham Mirfarah - National Cheng Kung University (Taiwan)
Abstract: In clusterwise regression modelling, it is assumed that data coming from several hidden clusters have different regression links. Various methods are used to identify observations' membership and subsequently to estimate each regression parameter. One of the widely used statistical frameworks for analyzing, clustering and classification purposes is a Mixture of linear Expert (MoE) models. Compared to the finite mixture regression models, the MoE models exploit the logistic function to allocate each observation to a specific group. This advantage of the MoE models enables us to use more information from the data and obtain an improvement in the data clustering. We introduced a robust MoE model for model-based clustering of the censored data with the scale-mixture of normal (SMN)distributional assumption on the unobserved error terms. An analytical expectation-maximization (EM) type algorithm is developed to obtain the maximum likelihood parameter estimates. The performance, effectiveness, and robustness of the proposed methodology are illustrated by conducting various simulation studies and analyzing a real-world dataset.
