- 單位 : 統計學系
- 分類 :
- 點閱 : 337
- 日期 : 2022-11-01
主講人：Mehrdad Naderi 博士 (Institute of Statistics, National Chung Hsing University)
講 題：Model-based clustering of skew three-way data: a novel methodology based on the mean-mixture of matrix-variate normal distributions
With the rapid growth of computer technology during the last two decades, the interest in statistical methods for analyzing data with a massive amount of information has been receiving worldwide attention. Over the past few years, the three-way (matrix- variate) data analysis is one of the burgeoning topics that inspire statisticians to attempt at developing new analytical methods. The term ``three-way" or ``matrix-variate" data refers to an array of realizations characterized by three dimensions: variables (rows), situations/locations (columns) and individuals (layers). For instance, in large-scale longitudinal studies, imagine n blood tests (columns) of N individuals (layers) where for each test p characters (rows) are measured.
This work introduces a unified finite mixture model based on the mean-mixture of matrix-variate normal distributions for modelling and clustering heterogeneous, asymmetric and leptokurtic three-way data. A statistical inference based on the likelihood function is proposed wherein the maximum likelihood (ML) parameter estimates are carried out by developing a computationally feasible EM-based algorithm. We conduct simulations to investigate the asymptotic properties of the ML estimators, classification assessment, and the capability of the Bayesian information criterion for model selection. We finally demonstrate the utility of the proposed methodology by analyzing a real-world data example.