Previous Chapter Chapter 65: Non-Parametric Measures of Association Next Chapter
Chapter 65: Non-Parametric Measures of Association
In situations in which it cannot be assumed that data are metric (interval- or ratio-scaled; see Measurement Scales entry), a non-parametric correlation technique may be utilized as a substitute for the Pearson correlation technique (see Correlation Analysis in the Dictionary).
A group of correlation measures deals with rank order data. Suppose two groups of consumers are asked to rank, in order of preference, the brands of a product class. The researcher wishes to determine the agreement, or correlation, between the two groups. Two possible statistics can be computed to accomplish this purpose: the Spearman rank order correlation coefficient, rs (rho), and the Kendall rank correlation coefficient, τ (tau).
The Spearman rank order correlation coefficient (also defined as the product-moment ...