3
$\begingroup$

How to tell if a specific cell from my 2x2 (also 2x2x2) contingency table (constructed for computing Chi-Square test of independence) is in favor or against the null hypothesis?

$\endgroup$
3
  • 1
    $\begingroup$ Make a table of residuals and look at that. If the table is high-dimensional, I would complementing with a correspondence analysis. Or, if you are looking for a few extreme cells in very large contingency tables, google for "Identifying extreme cells in a sizable contingency table" (without the quotes). $\endgroup$ Commented Feb 8, 2017 at 15:44
  • $\begingroup$ @kjetilbhalvorsen Thanks for the quick reply! I found this biostat.umn.edu/~dipankar/bmtry711.11/lecture_10.pdf (page 7/30) However, it only talks about a "rule of thumb" (page 8/30), in contrast I am looking for some statistical method to infer if the cell is in favor or against the null hypothesis. $\endgroup$ Commented Feb 9, 2017 at 10:42
  • $\begingroup$ I have finally found the answer. Page 2 (Calculating Residuals) of pareonline.net/getvn.asp?v=20&n=8 $\endgroup$ Commented Feb 9, 2017 at 11:16

1 Answer 1

2
$\begingroup$

Make a table of residuals and look at that. If the table is high-dimensional, I would complementing with a correspondence analysis. Or, if you are looking for a few extreme cells in very large contingency tables, google for "Identifying extreme cells in a sizable contingency table" (without the quotes). For example jstor: Detecting Outlying Cells in Two-Way Contingency Tables Via Backwards-Stepping.

For an example of such residuals see Interpreting residuals in chi-squared test

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.