Deriving Scale Values from the Analysis of Variance

Abstract

Subjects' responses to ordered stimuli may be used to derive interstimulus scale values without additional experimental procedures. From the conventional data matrix used for analysis of variance, it is possible to maximize selected ratios of sums of squares in a fashion which is equivalent to factor analysis or the classical discriminant problem. Given the assumption of a linear stimulus/response relationship, the calculations are straightforward and provide valuable additional information.