What is Chaterjee's xi correlation:
- Measure of the dependency between 2 variables.
- Able to capture non-linear dependencies. Pearson/Spearman only work for linear trends (see below).
- Tool for inference (if offers p-values or confidence intervals).
How to use it:
x = linspace(0, 20, 1000);
y = cos(x) + 0.3*randn(1, 1000);
xi = xicor(x, y);
Post any bugs/ideas/comments as issues!
The difference between Pearson's (
References:
- Sourav Chatterjee, A New Coefficient of Correlation, Journal of the American Statistical Association, 116:536, 2009-2022, 2021. DOI:10.1080/01621459.2020.1758115
- XICOR R package: https://cran.r-project.org/web/packages/XICOR/index.html