“Spurious nonsignificance” in rank correlation

Abstract

Spearman’s rho and Kendall’s tau rank correlations are acknowledged as highly utilitarian for a discipline increasingly responsive to the ordinal properties of much of its measurements. Rho and tau are conceptually reviewed, and concern is noted for the possibility that these rank correlation coefficients could be nonsignificant when underlying variables are strongly associated in a nonlinear fashion. An example of “spurious nonsignificance” is provided in which exploratory techniques involving a scattergram and the one-sample runs test are used to discern the true relationship of variables. Emphasis is placed on avoiding possible loss or misinterpretation of information in many nonsignificant rank correlations. © 1983, The Psychonomic Society, Inc.. All rights reserved.