Multivariate regression modeling in integrative analysis via sparse regularization

TL;DR

This paper introduces a multivariate regression approach for integrative analysis of multiple datasets, employing sparse regularization and an efficient algorithm to improve variable selection and predictive performance.

Contribution

It proposes a novel multivariate regression model for integrative analysis using sparse regularization and develops a convergent algorithm based on the alternating direction method of multipliers.

Findings

Method outperforms single-dataset analysis in simulations

Effective variable and group selection achieved

Demonstrated on wastewater treatment data with microbe measurements

Abstract

The multivariate regression model basically offers the analysis of a single dataset with multiple responses. However, such a single-dataset analysis often leads to unsatisfactory results. Integrative analysis is an effective method to pool useful information from multiple independent datasets and provides better performance than single-dataset analysis. In this study, we propose a multivariate regression modeling in integrative analysis. The integration is achieved by sparse estimation that performs variable and group selection. Based on the idea of alternating direction method of multipliers, we develop its computational algorithm that enjoys the convergence property. The performance of the proposed method is demonstrated through Monte Carlo simulation and analyzing wastewater treatment data with microbe measurements.