I have significanct F value for my demographic variable (nationality) in an independent variable, but the Scheffe post hoc comparison does not show any significant difference between nationality? But my ANAVO show that it was a
Hello Yew,
It is quite possible that the omnibus test for a factor yields a statistically significant result whereas a conservative (high protection level) post hoc test does not.
There are two reasons:
1. the omnibus (F) test is sensitive to any departure from the null hypothesis: mu1 = mu2 = ... = muk, so that it is possible that the nature of the differences in a given data set is not pairwise in form, but exists across more complex groupings (e.g., [mu1 + mu2]/2 vs. [mu3 + mu4 + mu5]/3).
2. The Scheffe is the most conservative (and therefore, the least powerful) of the post hoc tests, and imposes a stricter criterion for a mean difference to be declared statistically significant than do other post hoc tests. In my experience, studies with sample sizes that are modest at best may well be the most susceptible to this apparent contradiction in outcomes.
Good luck with your work!
Hi @Yew Gen Ng
I recommend you to perform other mean comparison :
1. LSD' (least significant difference) test for single mean comparison , or
2. DMRT (duncan multiple range test) for multiple mean comparison
Those post hoc test are stronger and more sensitive than scheffe test.
Good luck !
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA