1

Given a dataset of n records, each record having two attributes A1 and A2.

  • A1 can be one of two categories: X or Y.
  • A2 can be one of 10 categories, C1 to C10.

Given the following hypothesis: If A1 is X, then A2 is statistically significant mostly C1.

The null-hypothesis is, that the categories C1 to C10 are equally distributed.

How would I test for this statistical significance? I've been trying to do a chi-square, but did come to the conclusion that this might not be the right approach, since I would be testing if there is a statistical significant correlation between A1 and A2. However I want to test if there is a statistical significance of A2=C1, for a given A1=X.

What would be the right statistical test for this?

partikel
  • 11
  • It sounds like only the measurements where A1=X are relevant to the hypothesis you want to test. You could look at the measurements where A1=X, then do a one proportion test on the proportion that are C1. Under the null hypothesis, the true proportion is P(A2=C1 | A1=X) = 10%. If you care about detecting any departure form uniformity, you could do a chi-square test (on the measurements where A1=X). – user51547 Jul 29 '23 at 18:31

0 Answers0