Abstract:Cluster-weighted models (CWMs) extend finite mixtures of regressions (FMRs) in order to allow the distribution of covariates to contribute to the clustering process. In a matrix-variate framework, the matrix-variate normal CWM has been recently introduced. However, problems may be encountered when data exhibit skewness or other deviations from normality in the responses, covariates or both. Thus, we introduce a family of 24 matrix-variate CWMs which are obtained by allowing both the responses and covariates to be modelled by using one of four existing skewed matrix-variate distributions or the matrix-variate normal distribution. Endowed with a greater flexibility, our matrix-variate CWMs are able to handle this kind of data in a more suitable manner. As a by-product, the four skewed matrix-variate FMRs are also introduced. Maximum likelihood parameter estimates are derived using an expectation-conditional maximization algorithm. Parameter recovery, classification assessment, and the capability of the Bayesian information criterion to detect the underlying groups are investigated using simulated data. Lastly, our matrix-variate CWMs, along with the matrix-variate normal CWM and matrix-variate FMRs, are applied to two real datasets for illustrative purposes.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to conduct clustering analysis more effectively in the framework of matrix variables when the data exhibits skewness or other non - normality characteristics. Specifically, the existing Matrix - Variate Cluster - Weighted Models (CWMs) are mainly based on the matrix - variate normal distribution, which may encounter problems when dealing with skewed data or data with outliers. Therefore, this paper introduces a new family of matrix - variate CWMs. These models allow the response variables and covariates to be modeled separately or simultaneously by four existing skewed matrix - variate distributions or the matrix - variate normal distribution. In this way, the new models can handle the skewness in the data more flexibly, thereby improving the clustering effect. The main contributions of the paper include: 1. **Model Extension**: Propose 24 new matrix - variate CWMs that can handle cases where the response variables and covariates are skewed simultaneously or separately. 2. **Parameter Estimation**: Use the Expectation - Conditional Maximization (ECM) algorithm to derive maximum - likelihood parameter estimates. 3. **Performance Evaluation**: Evaluate parameter recovery, classification performance, and the ability of the Bayesian Information Criterion (BIC) to detect latent group structures through simulated data. 4. **Practical Application**: Apply the proposed models to two real - world datasets to demonstrate their application effectiveness in practical problems. Through these methods, the paper aims to provide a more flexible and powerful tool for handling clustering and classification problems in complex data structures.

Model-based clustering via skewed matrix-variate cluster-weighted models

Model-based clustering and classification using mixtures of multivariate skewed power exponential distributions

Model-based Clustering with Sparse Covariance Matrices

Parsimonious Seemingly Unrelated Contaminated Normal Cluster-Weighted Models

Mixtures of skewed matrix variate bilinear factor analyzers

A Bayesian approach for clustering skewed data using mixtures of multivariate normal-inverse Gaussian distributions

Robust Classification via Finite Mixtures of Matrix Variate Skew-t Distributions

Mixtures of regressions using matrix-variate heavy-tailed distributions

Assessing an Alternative for `Negative Variance Components': A Gentle Introduction to Bayesian Covariance Structure Modelling for Negative Associations Among Patients with Personalized Treatments

Clustering using skewed multivariate heavy tailed distributions with flexible tail behaviour

Bayesian mixtures of common factor analyzers: Model, variational inference, and applications

An EM algorithm for estimating the parameters of the multivariate skew-normal distribution with censored responses

Multivariate Matérn Models -- A Spectral Approach

Clustering with the multivariate normal inverse Gaussian distribution

Canonical fundamental skew-t linear mixed models

Model-based clustering based on sparse finite Gaussian mixtures

Finite mixture of regression models for censored data based on the skew-t distribution

FiMReSt: Finite Mixture of Multivariate Regulated Skew-t Kernels -- A Flexible Probabilistic Model for Multi-Clustered Data with Asymmetrically-Scattered Non-Gaussian Kernels

The orthogonal skew model: computationally efficient multivariate skew-normal and skew-t distributions with applications to model-based clustering

Scale mixtures of multivariate centered skew-normal distributions

Infinite mixtures of multivariate normal-inverse Gaussian distributions for clustering of skewed data