transDA: an R Package for Transformation Discriminant Analysis

J. Li${}^1$ and Y. Melnykov${}^1$

${}^1$ University of Alabama, Tuscaloosa AL 35487, USA

The Discriminant Analysis, which includes Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) [1], and Mixture Discriminant Analysis (MDA) [2], is recognized for its versatile and reliable approach to classification tasks, and it has countless applications across various fields of study. However, its effectiveness is often constrained by the assumption that each group or subgroup follows a Gaussian distribution, which may not hold for real-world data. Our R package transDA addresses this limitation by integrating transformation into discriminant analysis, allowing for skewness within data groups or subgroups. transDA can handle both non-transformation methods such as LDA, QDA, and MDA and transformation methods [3, 4]. This paper provides a solid theoretical foundation about transfortamion models and detailed descriptions of the functions of transDA, along with illustrative example.

Keywords: Discriminant analysis, R package, Transformation models