Application of Rasch analysis in sensory difference testing

Traditional discrimination methods either provide holistic product difference scores or focus on specific sensory attributes, often requiring multiple tests to capture both qualitative and quantitative insights. While aggregate-based analyses like ANOVA can statistically adjust product comparisons for assessor effects, they do not identify which individual assessors exhibit problematic rating behaviours, such as using limited parts of the scales or being too lenient or severe. Obtaining these diagnostic insights to guide targeted interventions (e.g., retraining or panel refinement) requires separate analyses that are not integrated into the standard discrimination testing framework.
This research explores the application of a Many-Facet Rasch Model (MFRM) as a diagnostic and analytical tool in sensory difference testing. MFRM addresses these challenges by estimating a single latent measure of overall product difference from combined ratings of multiple attributes, while simultaneously adjusting for individual differences in scale use. It also offers integrated quality control metrics that support panel diagnostics and highlight the discriminative value of individual attributes.
Across three studies, trained and untrained panels evaluated the intensity of various sensory attributes in three different food products. Rasch-derived overall difference measures aligned closely with results from the Difference-from-Control (DFC) overall difference test. Wright maps visualised the relative difficulty of perceiving attributes and the rating tendencies of individual assessors, while fit statistics and residual analyses revealed the contributions of individual attributes to perceived product differences and systematic rating patterns. MFRM further identified distinct types of individual scale-use bias, supporting targeted assessor training.
This study establishes the MFRM as a scalable, more insightful approach for sensory data analysis, with applications in quality control, product development, and panel management. Further research is encouraged to explore its utility across broader sensory and consumer testing contexts.