1
$\begingroup$

I have two multinomial data samples that both have N discrete categories. I know that a Kolmogorov-Smirnov test will let me know if the distributions of the two samples are different. But how can I statistically test if a particular category is significantly over-represented in one sample compared to the other?

$\endgroup$
  • 1
    $\begingroup$ Is the category identified before looking at the data? Or do you mean 'in any category'? $\endgroup$ – Glen_b Mar 20 '14 at 2:54
1
$\begingroup$

If you're looking for this for a specific individual category then the test simplifies down quite a bit. You can think of the data as being binary where 1=in that cateogry and 0=not in that category. Then the number in that category for each sample will be a binomial distribution with parameter $p$ and your test is a simple hypothesis test that $p$ is equal for both samples.

Does that make sense?

| cite | 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.