If there is a population that we consider fixed (finite) with size N - say the customers at a bank at the end of a given month - and we want to conduct an experiment on this population by randomly sampling m for treatment 1 and k for treatment2 (where m+k <=N), how can you test the difference in proportions between the two treatments?

enter image description here

1) Traditional tests for two proportions (e.g. z test, chi square) all assume infinite populations in their calculation of p-values correct?

Couple of thoughts - do any of these work in the case described?

  • Fishers Exact test?
  • Permutation tests?
  • Bootstrap?
  • 1
    $\begingroup$ Your actual population is infinite. It is not the people who came to the bank it is the theoretical class of bank customers. You are just dividing your sample from that population. For this you want Fisher's Exact Test. $\endgroup$ – Keith Jul 25 '14 at 13:25

Your Answer

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

Browse other questions tagged or ask your own question.