Multiple Comparison Tests Flashcards
To determine where difference is after using ANOVA
Also called “pairwise comparisons”
Multiple Comparison Tests
Performed after ANOVA (only if you find significance)
Most common
Test every difference, therefor are “exploratory”
Post-hoc
Performed instead of ANOVA (a priori)
Focused only on specific comparisons
You won’t see this used very often
Planned Comparisons
Divide alpha by number of statistical tests
Eg. alpha/6 = 0.05/6 = 0.008
Therefore use alpha = 0.008 for each post-hoc test
Downside is it is going to be very difficult to find significance with the alpha level that low
Bonferroni Correction
Essentially unadjusted t-tests
Fisher’s Least Significant Difference
Adjusted, but good balance of type 1 and type 2 error protection
Most commonly used
Sidak
Divides alpha by number of comparisons
Bonferroni Correction