A New Transformation to Reduce Skewness in Data with Bounded Support

Abstract

Normality is a critical assumption for many statistical analyses, yet real-world data often deviate from this requirement. To address this, statisticians may employ robust methods or apply transformations to achieve approximate normality. This study proposes a novel transformation for data from bounded-support distributions, such as the beta family, to induce symmetry or near-symmetry. The method utilizes a data-driven parameter, estimated through a simple and efficient procedure. Examples demonstrate the transformation's effectiveness, and its straightforward implementation allows easy reversion to the original data scale. We recommend this approach for statistical analyses requiring normality, even for datasets that are approximately normal, due to their versatility and robustness.