Rank-Based Item-Level Transformation for Harmonizing Multi-Scale Likert Data: A Comparative Evaluation

Abstract

Integrating Likert-type data from multiple instruments with heterogeneous response scales presents a persistent methodological challenge in applied multivariable modeling contexts involving ordinal data. This quantitative secondary analysis evaluates a rank-based transformation procedure, the Rank-Based Harmonization Framework (RBHF), alongside five benchmark transformations applied at the item level prior to aggregation. Methods were compared using descriptive distributional statistics, correlational analysis, multivariable linear regression, and residual diagnostics. Results indicated measurable differences across transformation strategies. Relative to benchmark methods, RBHF produced construct scores exhibiting closer alignment with normality indicators, smaller deviations from baseline correlation structures, comparatively larger adjusted R² values, and comparatively stable residual variance patterns within the analytic conditions examined. These findings reflect observable variation in how transformation procedures influence distributional properties, relational fidelity, and regression diagnostics when harmonizing multi-scale ordinal data. The RBHF procedure implements a minimum-rank inverse normal transformation at the item level and may be useful in applied research contexts where heterogeneous Likert-type instruments must be harmonized for joint analysis within parametric modeling frameworks.