Mixture of linear experts model for censored data: A novel approach with scale-mixture of normal distributions

TL;DR

This paper introduces a robust mixture of linear experts model using scale-mixture of normal distributions to effectively handle censored data and outliers, improving inference and clustering accuracy.

Contribution

It proposes a novel robust MoE model with scale-mixture distributions for censored data, addressing outlier sensitivity and inference issues in classical models.

Findings

The model outperforms traditional MoE in robustness and accuracy.

Simulation studies confirm the effectiveness of the proposed approach.

Real data analysis demonstrates the model's practical superiority.

Abstract

The classical mixture of linear experts (MoE) model is one of the widespread statistical frameworks for modeling, classification, and clustering of data. Built on the normality assumption of the error terms for mathematical and computational convenience, the classical MoE model has two challenges: 1) it is sensitive to atypical observations and outliers, and 2) it might produce misleading inferential results for censored data. The paper is then aimed to resolve these two challenges, simultaneously, by proposing a novel robust MoE model for model-based clustering and discriminant censored data with the scale-mixture of normal class of distributions for the unobserved error terms. Based on this novel model, we develop an analytical expectation-maximization (EM) type algorithm to obtain the maximum likelihood parameter estimates. Simulation studies are carried out to examine the performance, effectiveness, and robustness of the proposed methodology. Finally, real data is used to illustrate the superiority of the new model.