THE EFFECT OF FACTOR RANGE ON WEIGHT AND SCALE VALUES IN A LINEAR AVERAGING MODEL

This study investigated the effect of variation of a factor's range on its psychological importance, or weight, in multiattribute judgment tasks. The influence of a factor's range on its weight is important to both the psychological theory of judgement and the practice of judgment analysis. If factor range affects importance, suboptimal preference reversals may occur, and standard methods of weight assessment may produce unstable estimates of psychological importance. Such findings would also have implications for the basic definition of importance, or weight, in multiattribute judgment tasks. Previous studies of this topic have been limited either to tasks involving equivalent judgment factors, to methods of analysis that confound the influence of range on weights and scale values, or to between-subject designs.;Analyses of both the rank orders and the ratings showed a variable, but generally positive, effect of factor range on weight, and a positive, though less extensive, effect on the subjective distance between factor levels. The strongest influence was found for the perceptual task, and for the judgments generated by the models fit to subjects' ratings for the social policy task.;These results suggest that the concept of weight is not psychologically independent of range. Judgment researchers can therefore expect that varying the range of judgment factors may result in both direct and predicted preference reversals, and that the measured importance of factors will be dependent upon range, even when measurement methods are used that control for the statistical confounding of weights and scale values.;The purpose of this study was to test the hypothesis that a factor's range has a positive effect on its psychological importance, independent of any influence on the factor's subjective scale values. A within-subject design with fifty-five subjects was employed with three two-factor judgement tasks: a perceptual task, an apartment evaluation task, and a social policy task. Subjects both rank ordered and rated two sets of alternatives for each task--each set containing alternately a large range on one factor and a small range on the other. Ratings of factors taken singly were also made to enable unconfounded estimation of weights and scale values.