Calculating and Reporting Effect Size for Fisher's Exact Test?

[nporot]

Hi everyone,

I am writing a paper for publication. This is my first time doing a survey study, and also my first time working with categorical data (I, like many people unfortunately, have learned statistics bit by bit, as I have needed it in my research). The journal I am submitting to requests reports of effect size. I mostly used Pearson's chi-square for my comparisons, but in some cases, my 2 x 2 contingency tables had low expected values, and so I used Fisher's exact test. For the chi-square tests, I calculated the phi coefficient as a measure of effect size. For Fisher's exact test, I have read that odds ratio is an acceptable effect-size measure. Two questions, then:

(1) Can I report phi for some of the tests in my study and the odds ratio for others, or should I have a unique measure of effect size for all of the tests in my study?
(2) How do I report an odds ratio as a measure of effect size in my paper? (I have not been able to find this information in any APA format guidelines online).

Thanks in advance

 

[hlsmith]

I would run Fisher's exact for all tests, it is preferred and unless you have huge samples it should be pretty quick. Then report all tests using odd ratios plus confidence intervals.


Female respondents had 5.1 (95% CI: 2.5, 6.0) times greater odds for outcome X than male respondents.

 

[nporot]

Thanks very much for the helpful reply!

I would run Fisher's exact for all tests, it is preferred and unless you have huge samples it should be pretty quick. Then report all tests using odd ratios plus confidence intervals.


Female respondents had 5.1 (95% CI: 2.5, 6.0) times greater odds for outcome X than male respondents.