0

I have 2 populations which I want to test if the proportions between group A and B are the same for each location or not by using hypothesis testing. It's very much as this problem described here: Comparing multiple sample proportions between two groups. However in my case it's not between two independent groups but group B population is a subset of group A's. Can I still use the method described in the link or do I have to think differently? Is there a fallacy I am missing out here? My data looks like this:

                           Group A    Group B
Location 1                 20090      17
Location 2                 24604      10
Location 3                 12284      5
Location 4                 14843      12
Location 5                 20592      9
Location 6                 8715       4

I am thinking because I have such a low count in group B, Fisher's exact test should be better suited here if I am gonna use something similar as the link.

arezaie
  • 157
  • 9
  • If these are truly *populations* then just compare the proportions: they are whatever they are. If they are *samples,* then just compare group $B$ to everything that's in $A$ but not in $B.$ The chi-squared test will work fine. – whuber Apr 25 '21 at 13:59
  • Please clarify what the proportions are in reference to...and to make sure I'm understanding the question, the last location has 4 B, 8715 A, and thus 8711 A-which-are-not-B...¿is this correct? – Gregg H Apr 25 '21 at 17:06
  • @GreggH yes that is correct. – arezaie Apr 26 '21 at 08:33
  • 1
    And you are just asking if the distribution across locations is the same as for B as for A-not-B. If so, this is just a multinomial comparison. – Gregg H Apr 26 '21 at 21:31
  • Yeah, after your input I realize I could do a Fischer exact test with simulated p-values in R. It would give me a p-value which meant I couldn't reject H_0 – arezaie Apr 27 '21 at 06:40

0 Answers0