A Comment on Replicated Paired Comparisons
Timothy Paul Hutchinson · Journal of the Royal Statistical Society Series C (Applied Statistics) · 1979
SUMMARY Models for the pattern of responses in replicated paired comparison tests are discussed. It is suggested that stochastic models that permit preference to vary continuously over the population, as suggested by Wierenga (1974), are more in line with modern psychological research than are those which sharply divide the population into distinct groups, as do those of Horsnell (1977). Modifications to Wierenga's model that meet Horsnell's objections are suggested. The resulting models give satisfactory fits (P>O 1) to 11 of the 12 datasets used by Wierenga (1974), and to 4 of the 6 datasets used by Horsnell (1969).