In many examples throughout the social sciences, the strength of relationship between a response variable Y and a covariate X differs for different values of the covariate. Usual measures of association such as the correlation coefficient are unable to capture this varying strength of relationship. The correlation curve introduced by Bjerve and Doksum (1993) measures local association by measuring the local variance explained by regression. In this paper we explain the construction of the correlation curve, focusing on the concept of variance explained by regression, and we show how the curve is a useful data-analytic tool in a variety of examples. Various methods for estimating the correlation curve are illustrated. The paper is designed to facilitate application of the curve to data sets by practitioners in quantitative disciplines.

>