Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Measuring and detecting associations: Methods
based on robust regression estimators or
smoothers that allow curvature
Rand R. Wilcox*
Department of Psychology, University of Southern California, Los Angeles,
California, USA
This paper considers the problem of estimating the overall strength of an association,
including situations where there is curvature. The general strategy is to fit a robust
regression line, or some type of smoother that allows curvature, and then use a robust
analogue of explanatory power, say h
2
. When the regression surface is a plane, an
estimate of h
2
via the Theil–Sen estimator is found to perform well, relative to some
other robust regression estimators, in terms of mean squared error and bias. When
there is curvature, a generalization of a kernel estimator derived by Fan performs
relatively well, but two alternative smoothers have certain practical advantages. When
h
2
is approximately equal to zero, estimation using smoothers has relatively high bias.
A variation of h
2
is suggested for dealing with this problem. Methods for testing
H
0
: h
2
¼ 0 are examined that are based in part on smoothers. Two methods are found
that control Type I error probabilities reasonably well in simulations. Software for
applying the more successful methods is provided.
1. Introduction
As is evident, measuring and detecting an association between two or more random
variables is of fundamental importance. For two random variables, say X and Y , certainly
the most commonly used measure of the strength of the association is Pearson’s
correlation, r. And the most common strategy when attempting to establish
dependence is to test
H
0
: r ¼ 0: ð1Þ
And of course, when dealing with p predictors, there is the usual squared multiple
correlation coefficient, which arises quite naturally when using least squares regression.
*Correspondence should be addressed to Professor Rand R. Wilcox, Department of Psychology, University of
Southern California, Los Angeles, CA 90089, USA (e-mail: rwilcox@usc.edu).
The
British
Psychological
Society
379
British Journal of Mathematical and Statistical Psychology (2010), 63, 379–393
q 2010 The British Psychological Society
www.bpsjournals.co.uk
DOI:10.1348/000711009X467618