[An empirical comparison of 5 standard procedures for one-dimensional scaling].

Abstract

Scales on the badness of 10 offenses/crimes are derived by using the five most common scaling methods: averaging of the badness-ratings of the stimuli on an 11-point scale; mean ranks of complete rankings of the 10 stimuli; Thurstone scales for complete forced-choice and constant-sum pair comparisons, resp.; and ratio scales derived from pairwise ratio judgments for all pairs. From older investigations, a logarithmic correspondence of ratio scales and Thurstone or average ratings, resp., can be predicted. The data corroborate this prediction. They show furthermore that the ratio scale has an exceedingly close, logarithmic relation to each of the other four scales. We therefore conclude that, once a universe has been shown to be scalable, a scale can be constructed very economicaly by simply averaging ratings.