An Informational Measure of Correlation

Summary Informational considerations lead to a natural generalization of the classical correlation coefficient of a normal distribution. The generalized coefficient, here called the informational coefficient of correlation , is a function of the joint probability density distribution p ( x, y ) of the two variables x and y , is invariant under a change of parameterization x′ = f ( x ), y′ = g ( y ), and reduces to the classical correlation coefficient when p ( x, y ) is normal.