Hello,
I am running an ANOVA between 3 groups (Group 1 N=148, Group 2 N=147, Group 3- N=144; equal variances assumption met) .
I am having difficulties in deciding upon the most adequate post-hoc test (Bonferroni vs. Tukey) and the available information is mixed.
Can someone please help me with a scientific explanation regarding the adequate post-hoc test for my data? Thank you !
For 3 groups (so there are three possible comparison) it does not really matter. You get very similar p-values with both procedures. Tukey's HSD will be very slightly more powerful.
Tukey HSD is the commonly recommended test by most statistical softwares. The Bonferroni is a very robust test, sometimes very discriminative, mostly employed where Tukey test cannot be used by reason of failure to meet some assumptions. Please certify that your data meets critical assumptions for using the Tukey HSD especially equality of variance. You can verify equal variance using something like Levene's Test or the Brown-Forsyth test, which are available in most statistical software packages. If neither of these is available, then you could try Hartley's Test.
The detailed answer is that the Tukey HSD is a proper "post hoc" test whereas the Bonferroni test is for planned comparisons. The Bonferroni test also tends to be overly conservative, which reduces its statistical power. ... Should your data *not* have equal variance, then there are other posthoc tests that might be used.
Hello Alexandra Calugar. With k = 3 groups, and with homogeneity of variance being a reasonable and defensible assumption, I would use Fisher's LSD, as it will be more powerful than the alternatives.
https://www.researchgate.net/post/When_do_we_use_the_Fisher_LSD_significance_with_one-way_ANOVA_test
HTH.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running