Some problems in testing the double cancellation condition in conjoint measurement (Q1108727)

Two problems involved in testing the conjoint measurement condition of double cancellation are analysed. 1. It is argued that in the case of complete data matrices tests of double cancellation either confirm or falsify the hypothesis of additivity and that, contrary to popular belief, so-called ``no-tests'' are really acceptances. 2. It is shown that with a weak order upon \(3\times 3\) matrices both independence and a single test of double cancellation (which is logically independent of independence) are necessary and sufficient for additivity. Hence, if a weak order satisfies independence in such matrices only this single test of double cancellation needs to be performed.