A psychological scaling model for testing order hypotheses
Mark L. Davison · British Journal of Mathematical and Statistical Psychology · 1980
This paper describes a scaling model which is based on Goodman's concept of quasi‐independence and which can be used to test an a priori order hypothesis for dichotomous or rank order data. The order hypothesis may be unidimensional or multidimensional. Methods based on the model provide a means of investigating Guttman scales, Thurstone scales, developmental stage sequences, first‐in‐first‐out developmental sequences and Coombs' unfolding theory for preferences. The methods are designed to assess whether or not there are dependencies among a set of response variables which go beyond those imposed by the experimental task and, if so, whether or not those dependencies can be accounted for by the hypothesized orderings of the variables. The methods are unique in that they are applicable to both ipsative and non‐ipsative data.