multiple comparisons - corrections

Question 1

what are corrections

Answer 1

methods designed to control for the number of tests

Question 2

what do corrections do

Answer 2

keep the type 1 error rate at the intended alpha level

Question 3

Bonferroni

Answer 3

for a small number of planned comparisons or pairwise comparisons
divides alpha by the number of tests
assumes comparisons are independent of one another
sacrifices more power than Tukey
can be applied to any set of contrasts or linear combinations

Question 4

Sidak

Answer 4

bit more powerful than bonferroni
assumes all comparisons are independent of one another
less common than Bonferroni
more difficult to calculate

Question 5

Scheffe

Answer 5

calculates the p-value from f distribution
most conservative (least powerful)
controls the family alpha level for testing all possible contrasts
should be used for unplanned contrasts
not used for testing pairs of treatments
makes the critical value of F larger for a fixed alpha depending on the number of tests