Chatterjee's Xi Coefficient
Audience: high-schoolundergraduategraduate
Tags: statistics
This blog post explores an elegant and surprisingly recent measure of correlation developed in 2019. We'll see the intuition for this new formula and how it compares with other measures of correlation we're more accustomed to in an introductory statistics class
Analytics
Comments
It took a bit of time to understand the flow of the presented arguments but once I understood them I thoroughly enjoyed the article. For someone like me, who deals with mathematical objects that are too complex to ever visualize faithfully, articles such as this one which use simple examples to explain a method are a joy to read.
This was a good quick reading for introducing the Xi coefficient, thanks! It was a bit more comprehensive than SciPy’s scipy.stats.chatterjeexi
documentation, and much easier to read than the original paper.
I think it would be nice to include a very concrete application to tell a story of Xi succeeding in a real-world situation where other coefficients fail.
nice read, the post could benefit from an interactive widget.
It would have been nice if you had explained the rank in a quick note, I had to look it up. This was a good explainer, though it could have been motivated more concretely. It basically assumed I already care about correlation and an application example that only works with the new measure would have been nice, also for memorability.
Cool topic! I would personally prefer some user interactivity, and also explanation of what rank is.
Minor nitpick but I was confused by the notational inconsistency of sometimes using and sometimes using for spearman’s. Overall I really liked this piece. It had a nice motivation and a clear direction; diagrams were nice. Generally a very clear writing style and I like the use of footnotes. It would have been awesome to have some sort of interactive element where you could plot points and see all three coefficients.
I understood nothing but had a splendid time. He explained exactly what was necessary: the intuition, and the application, even some of the history. 10/10, “What is considered a large value of d?” best quote.
Very novel, very memorable. As for motivation, yeah, a better one was discovered and the build up to it was very well orchestrated. Demonstrated previous formulas strengths first, then their weaknesses, each new one succeeding where the previous fell short.
For a written body of text, this didn’t bore me whatsoever. Huge.
A very clear description of how the equation came to be, as well as the steps for deriving it. I enjoyed learning about this a lot. The only thing that came to mind was that rank wasn’t clearly defined, and so some people might not be able to follow.
I was surprised how simple the intuition behind the formula is - often statistical measures can be a bit opaque at first but this was crystal clear