Comparing F-statistics

#1
Is it possible/appropriate to compare two or more F-statistics results, say from an ANOVA or MANOVA, to say which analysis has a larger difference when both p-values are the same(and sig)?

Example -
3 pair-wise MANOVAs:
λ 0.023 , F=45.541, p<0.0001
λ 0.023 , F=49.073, p<0.0001
λ 0.014 , F=57.554, p<0.0001

Is it appropriate to say that based on the F values, the third analysis has the largest variation between the groups?

Thanks!
 

obh

Active Member
#2
You shouldn't compare the p-values ...the p-value only "say" if the results are significant. even if the difference is very small but significant.

Did you think about comparing the effect size?
 

ondansetron

TS Contributor
#3
Is it possible/appropriate to compare two or more F-statistics results, say from an ANOVA or MANOVA, to say which analysis has a larger difference when both p-values are the same(and sig)?

Example -
3 pair-wise MANOVAs:
λ 0.023 , F=45.541, p<0.0001
λ 0.023 , F=49.073, p<0.0001
λ 0.014 , F=57.554, p<0.0001

Is it appropriate to say that based on the F values, the third analysis has the largest variation between the groups?

Thanks!
When doing a MANOVA, you want to consider which omnibus test might be best for your situation. It's possible one or more is most ideal based on it's performance characteristics, so you would look at the one/few you preselected. Do some research on when to use Pillai's trace vs Wilk's Lambda and so on and see which is best for your scenario.