1
$\begingroup$

I have seen the questions like Correlations with unordered categorical variables , but I cannot implement the chi-square test since I don't have a chi-square table representing joint distribution. Besides, It was said that distance metrics like Euclidean distance is not reasonable. Currently, I just have several 1*n data, which are discrete variables or I can use their Gumbel softmax distribution approximation.

How can I test the association of those unordered category distributions? PS: For example, one data is [0.1 0.5 0.3 0.1], another one is [0.2 0.5 0.1 0.2]. Actually, they are quite similar if we sort them and compare them.

$\endgroup$
0
$\begingroup$

Personally, I decided to use the JSD divergence.

$\endgroup$

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.