I have a new article up on the Win-Vector blog, about the ξ ('XI') correlation coeffient that was recently introduced by Professor Sourav Chatterjee in his paper, “A New Coefficient of Correlation”.
Unlike traditional correlation coefficients, ξ does not assume that the relationship between x and y is linear; in principle, it can be any functional relationship. While this is potentially useful, ξ also has some disadvantages, too.
In my Win-Vector post, I run some informal experiments to try to get a sense of what different values of ξ might mean.
Please check it out.