p value for Conover's test
Hi everyone,
I would like to know when using Conover's test as the post hoc comparison for Friedman's test, which p-value I should use? The output table has 3 different p values, namely, p, pbonf, and pholm. And also, is the alpla level automatically adjusted when performing Conover's test in JASP? Thanks in advance!
- I have included a screenshot below. If I set an alpha level at 0.05, and I have a total of 10 pairwise comparisons performed by Conover's test, which ones are significantly different based on the output below?
Comments
This depends on whether you wish to correct for multiplicity, and what method of multiplicity-correction you prefer, given the research context at hand. Usually Bonferonni is considered very strict. Your best bet is to consult the background literature on this tricky issue.
Cheers,
E.J.
Hi @EJ,
Thanks for your response!
I have decided to do a p-value correction due to multiple testing. If I want to use Bonferroni correction, given that I have 10 pairwise comparisons, the altered alpha level will be set at 0.05/10 = 0.005. Then based on the output of JASP, does that mean I should go to the column of pbonf and then find if there is any p-value <= 0.005 in order to decide it's statistically significant? Or, should I just use the original alpha level (0.05) as the threshold to compare with the pbonf column results and decide the result?
" should I just use the original alpha level (0.05) as the threshold to compare with the pbonf column results and decide the result"
Yes. You can see that, for instance, the .008 p-value corresponds to a .080 p-bonf value.
E.J.
@EJ
Got it ! Thank you very much!