Questions are raised concerning the appropriateness of statistical tests of agreement among judges. It is noted that, following Coombs, judges perform 2 kinds of tasks: Task A (preference) asks the ...judge to compare a set of stimulus objects to his own ideal set, & Task B (similarity) asks the judge to compare a set of stimulus objects in terms of each other. Assumptions of tests of signif of rank orders involve the assumption that all permutations are equally likely under chance, but in practice some permutations seem to be impossible or highly unlikely, thus seriously questioning the use of such tests of signif as indicators of agreement. Suggestions for further study are discussed. C. R. Shepherd.
If the normative data includes information on the mean intragroup profile correlations, the test user can discount the correlation of averages, making the correlation of the Ss profile with the ...average profile interpretable as an average correlation between individuals profiles.
The problem of this study is the examination of a high school group situation in which two broad background elements exist, the rural and the urbanwith the aim of seeking the statistical importance ...of these elements in their influence upon the peer acceptance within the group.
Statistics (3rd ed.) Russek, Estelle; Hays, William L.
Journal of the American Statistical Association,
03/1982, Letnik:
77, Številka:
377
Journal Article
A computationally convenient way to find average tau is pointed out, and the possibility is raised of simple chi-square tests of significance for average τ as well as Kendall's coefficient of ...agreement u. A method for obtaining the rank order of "best fit" to a group of rankings in terms of average tau is also discussed. Finally, an "analysis of agreement" is proposed based upon a partition of average tau into agreement within and between experimental groups.