Replicated paired comparisons at the individual level

Abstract

A model and generalized least squares (GLS) analysis is presented for paired comparisons replicated within individuals. This analysis is applicable to any pairwise judgement or datum that is interprétable as a preference or perceptual difference, and it permits any (unknown) pattern of correlations among the paired comparisons. Exact individual‐level hypothesis tests and GLS estimates are developed from Wishart distribution theory and applied to a consumer taste panel.