I am proposing this question for a chi-squared test of association between party and voting. The following table is based on how (R)-Republican and (D)-Democrats voted on a specific bill

     yes no
   R  99 11
   D   1 48

Since the (D, yes) cell is 1, could I be violating Cochran's rules for a chi-squared test?

  • $\begingroup$ What is the expected count for D-yes? $\endgroup$ – Alexis Dec 5 '17 at 22:48
  • $\begingroup$ 30.817 found from (100*49/159) $\endgroup$ – Nathgun Dec 5 '17 at 22:57
  • $\begingroup$ Isn't the rule that the $\chi^{2}$ contingency table test should not be used when expected counts in any cell are less than 5? $\endgroup$ – Alexis Dec 5 '17 at 23:04
  • $\begingroup$ @Alexis I should have specified that the table listed here is the observed counts. None of the expected count cells are less than 5 $\endgroup$ – Nathgun Dec 5 '17 at 23:28
  • 1
    $\begingroup$ Cochran's main point under this heading was that a rule such as all expected counts being at least 5 was over-cautious. $\endgroup$ – Nick Cox Dec 6 '17 at 2:35

Answer: contingency table $\chi^{2}$ tests with expected counts less than 5 were deemed inappropriate by Cochran, but your example is of observed counts.


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.