Partial Least Squares Regression for Binary Data

Classical Partial Least Squares Regression (PLSR) models were developed primarily for continuous data, allowing dimensionality reduction while preserving relationships between predictors and responses. However, their application to binary data is limited. This study introduces Binary Partial Least S...

Full description

Saved in:
Bibliographic Details
Main Authors: Laura Vicente-Gonzalez, Elisa Frutos-Bernal, Jose Luis Vicente-Villardon
Format: Article
Language:English
Published: MDPI AG 2025-01-01
Series:Mathematics
Subjects:
Online Access:https://www.mdpi.com/2227-7390/13/3/458
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Classical Partial Least Squares Regression (PLSR) models were developed primarily for continuous data, allowing dimensionality reduction while preserving relationships between predictors and responses. However, their application to binary data is limited. This study introduces Binary Partial Least Squares Regression (BPLSR), a novel extension of the PLSR methodology designed specifically for scenarios involving binary predictors and responses. BPLSR adapts the classical PLSR framework to handle the unique properties of binary datasets. A key feature of this approach is the introduction of a triplot representation that integrates logistic biplots. This visualization tool provides an intuitive interpretation of relationships between individuals and variables from both predictor and response matrices, enhancing the interpretability of binary data analysis. To illustrate the applicability and effectiveness of BPLSR, the method was applied to a real-world dataset of strains of <i>Colletotrichum graminicola</i>, a pathogenic fungus. The results demonstrated the ability of the method to represent binary relationships between predictors and responses, underscoring its potential as a robust analytical tool. This work extends the capabilities of traditional PLSR methods and provides a practical and versatile solution for binary data analysis with broad applications in diverse research areas.
ISSN:2227-7390