4
$\begingroup$

I use Cramer's V to calculate correlation of features in a dataset made of only nominal features.

Let's consider the following dataset:

a  |  b
--------
0  |  0
0  |  1
0  |  0
1  |  2
1  |  2
1  |  3

Calculating Cramer's V for features a and b yields 0.707. Since it's symmetric, there's information loss in this case - as we can see, knowing the value of b means we know for sure what is the value of a, but this is no the case if we are given the value of a; in this case, the number possible values of b decreases, but it's still not known for sure.

I'd like to find an asymmetric metric that will provide this information for nominal values - meaning, will give a different value when calculated a -> b and b -> a. Is there anything like this?

$\endgroup$
8
$\begingroup$

I found what I was looking for - it's called Theil's U, or the Uncertainty Coefficient.

I've used it in this Kaggle kernel, you can check it out for an example and code implementation in Python

EDIT: I also have a blogpost about it.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.