REPLICATED PAIRED COMPARISONS AT THE INDIVIDUAL-LEVEL

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 interpretable 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.