A Monte Carlo evaluation of four techniques for capturing and clustering rater strategies

Frequently judges or raters must make overall evaluations of multicue objects or situations along some dimension. Several studies have used simulated stimuli and multiple regression procedures to investigate the capturing and clustering of judgment strategies (Dudycha & Naylor, 1966; Naylor, Dudycha, & Schenck, 1967; Naylor & Wherry, 1965; Wiggins, Hoffman, & Taber, 1969; Madden, Note 1). A rater strategy is essentially the way the rater (or judge) utilizes the available cues in evaluating the set of stimuli. The distinction between capturing and clustering strategies has been made clearly in the literature (e.g., Dudycha, 1970). "Capturing" a strategy refers to the extent to which a rater's evaluations are predictable from the given characteristics of the stimuli. This is usually indexed as the squared multiple correlation between the stimuli characteristics and overall evaluations. "Clustering" concerns the grouping of raters (or judges) on the basis of the similarities that exist between their strategies. While multiple regression has been commonly used for "capturing" strategies, a number of procedures have been reported for clustering raters with similar strategies. Each of the procedures has been used in empirical studies reported in the literature, but there has been only one study (Dudycha, 1970) which reported a Monte Carlo evaluation of any of these procedures. The purpose of the present study was to investigate the effects of three parameters on the effectiveness of four clustering procedures in discriminating predetermined strategy groups.