Polynomial psychophysics of group risk perception

The Coombs and Huang (1970) distributive theory of perceived risk is reinterpreted as a more robust statistical hypothesis to describe central tendencies of noisy replicates drawn from a homogeneous population. Barron's (1976) sample of 13 business faculty rank-order responses are pooled to obtain a replicated complete 3 × 3 × 3 design which is analyzed by a new stochastic conjoint measurement (SCJM) approach to axiomatic data analysis. SCJM implements statistical analogues of the deterministic Krantz and Tversky (1971) diagnostics for error-free data. SCJM diagnosis based on a series of one-sided nonparametric two-cell comparisons at the α = 0.04 level supports the hypothesis of interaction between the expected-value and number-of-plays attributes of gambles yet contradicts Barron's odd-even effects hypothesis. SCJM diagnosis with two-cell α < 0.04 supports an additive statistical model.