A Distributional Measure of Correlation

by Dror G. Feitelson.

Abstract: Common metrics of correlation are based on the degree to which the value of one random variable can be used to predict the value of another random variable. However, there are forms of correlation that do not lead to a high ability to predict actual values. We identify such a correlation, and suggest it can be measured by a certain monotonicity property among distributions of the dependent variable, for different ranges of the independent variable.

Key Words: correlation coefficient, data analysis, distribution prediction

Dror G. Feitelson, feit@cs.huji.ac.il

Editor: Richard G. Graf,rgraf@sunstroke.sdsu.edu

READING THE ARTICLE: You can read the article in portable document (.pdf) format (2051283 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 4224 times since July 24, 2006.

Return to the InterStat Home Page.