A CHANGEABLE SCALE VALUE MODEL FOR RANK ORDER DATA

Abstract

A generalized version of Takane and Carroll's (1981) (Psychometrika, 46, 389-405) scaling model is developed to analyze rank order data. This model assigns to each stimulus n (the number of stimuli) different scale values each of which is to be utilized at its respective occasion of successive first choice in the ranking task. The n scale values assigned to a stimulus is predicted by using the cumulative binominal distribution function for an assumed set of n-1 Bernoulli trials of a simple judgment about the same attribute of that stimulus as is to be scaled from the ranking data. Examples of the application of the model are reported. They have demonstrated that the generalized model can account for all the data sets examined better than the original one.