5 ms·
Is there a way to extend this to measure a multivariate relationship? For example: Cor(X, Y & Z) I know you could run them pairwise but it’s possible Cor(X, Y
by laplacesdemon48 5y ago
Is there a way to extend this to measure a multivariate relationship?
For example: Cor(X, Y & Z)
I know you could run them pairwise but it’s possible Cor(X, Y) and Cor(X, Z) are close to zero but Cor(X, Y & Z) is close to 1.
- kortex 5y agoSeems easy enough to play around with, and need not be strictly numbers either, as long as rank is defined on your fields (they are sortable...total ordering? I'm rusty on my terminology). Basically sort X, take the variance of the rank of Y, Z, etc. Much in the same way you would compute multi variable correlation.
- tomrod 5y agoThis is a good question. When I read the article I'll keep it in mind!