Post Hoc Comparisons/tests Flashcards
What are the purposes of all tests?
1) indicate where the mean differences lie
2) Maintain alphafamilywise at some predetermined level
What can you tell if omnibus F is significant?
That there will be at lest one significant post-hoc comparison. However, none of the comparisons that you actually conduct may be significant.
What is Tukey also called?
HSD (Honestly Significant Difference)
When do we use Tukey (HSD)?
- When making all possible pairwise comparisons.
- Use when all possible pairwise comparisons
- Controls over all alpha when you have homogeneity of variance.
What is Tukey (HSD) based on?
- Based on a stat called the standardized range statistic (q)
What is the “easy” way to perform test?
1) calculate differences between all pairs of means (Xbars)
2) Compare these differences to a minimum, or critical difference between means
What does the critical difference equal?
HSD
What is the formula for HSD?
q√MSw/n
MSw comes from omnibus F (in ANOVA)
How do you find q?
in the q table, it is a tabled value.
How do construct the table for Tukey?
Rank order the means by going highest to lowest (one row going left to right, the other column going from top to bottom).
Then subtract the means, ignoring the signs.
How do you get the df for error term? (to find q in q table)
MSw df, so N-K
Once you have your HSD (q√MSw/n, how do you find which means are significant?
Go back to the table, find means that are greater than HSD
What do you do if n is unequal in a post hoc test?
Use the harmonic mean for each PAIR of groups
This is called the Tukey-Kramer procedure
