Assessing Potential Heteroscedasticity in Psychological Data: A GAMLSS approach

Abstract

This paper provides a tutorial for analyzing psychological research data with GAMLSS, an R package that uses the family of generalized additive models for location, scale, and shape. These models extend the capacities of traditional parametric and non-parametric tools that primarily rely on the first moment of the statistical distribution. When psychological data fails the assumption of homoscedasticity, the GAMLSS approach might yield less biased estimates while offering more insights about the data when considering sources of heteroscedasticity. The supplemental material and data help newcomers understand the implementation of this approach in a straightforward step-by-step procedure.