An Introduction to Intraclass Correlation that （介绍组内相关）.pdf

An Introduction to Intraclass Correlation that Resolves Some Common Confusions Peter J. Taylor Programs in Science, Technology Values, Critical Creative Thinking, and Public Policy University of Massachusetts, Boston, MA 02125, USA 617 287 7636; 617 287 7656 (fax); peter.taylor@ KEY WORDS: Analysis of Variance; Double-entered data; Intraclass correlation; Linear Correlation; Principal Component Analysis; Standardized variables Abstract This article resolves common confusions about intraclass correlations in a way that starts from Weldons (2000) introduction of the usual linear correlation and regression in terms of standardized variables and minimum squared distances, but also makes use of principal component analysis and refers to analysis of variance. The intraclass correlation for a set of classes in which the order of the values is arbitrary (and independent from one class to the next) is a non-negative quantity. Negative estimates are possible and can be interpreted as indicating that the true intraclass correlation is low, that is, two members chosen randomly from any class vary almost as much as any two randomly chosen members of the whole population. 1. Introduction Intraclass correlation (hereon: ICC) refers to a number of quantities, but the simplest form is the usual linear (Pearson product-moment) correlation among a set of pairs of values when the order in each pair is arbitrary. That would be the case if we wanted to know the correlation of heights in same sex couples or the agreement of two independent ratings of an exam done by a set of students (where the two raters differed from one student to the next). This form of ICC—which is the subject of this article—can be calculated after a double-entry of the set of pairs, where each is included twice in the calculation of the linear correlation: once in one order; once in the other. The double-entry method can be generalized to classes of size k, where each of the k(k- 1)