THE INTERPRETATION OF MEASURES OF PROFILE (DIS-) SIMILARITY
Abstract
Empirical and mathematical results were presented which strongly suggest that the generality observed for D2 is not a truly dyadic effect. Instead this generality appears to arise from consistencies in a rater's use of items apparent in his description of even a single Figure. One such consistency is the rater's tendency to use extremes in description. The cluster difference, d, and the q-correlation, R, were tentatively suggested as appropriate dyadic statistics for interpersonal perception research. Little advantage seemed to result from the use of a nonparametric measure of q-correlation, tau. The necessity of giving prior consideration to more easily interpretable monadic and linear combinations of item-ratings was stressed. Greater use of more conventional methods of multivariate statistical inference was urged. The complexity of the problem and the limitations of the above results, as well as some of the results themselves, guarantee that exceptions to any of the general rules proposed will be found in individual research situations. But this diversity clearly makes correct interpretation of such research dependent on greater exploration in depth than has been typical of multivariate interpersonal perception studies. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1961
- Accession Number
- AD0263337
Entities
People
- J. Mcv. Hunt
- Rolfe Laforge
Organizations
- University of Illinois Urbana–Champaign