Post hoc alternatives

Bonferroni

Nominal type I error rate for each comparison is alphaFW divided by the nb of comparisons

Sidak

Adjusts type I by the no of testS but less conservative than bonferroni

Dunnet

Used when compare one group (usually control) to k-1 groups No pairwise comparisons, requires f crit from specialized table

Holm

Sequential mean comparison with a readjustment to the nominal type I error using the bonferoni correction

Fisher Hayter

Following a significant omnibus f test, compute a minimum absolute difference between means required for statistical significance based on the Q crit with k-1 dis (1 less df than for Turkey) and compare the two most discrepant means first (less conservative than tukey)

Newman Keuls

Start with the largest mean diff, if Ho is rejected the test proceeds until it is retained just like fisher hayter. The minimum absolute difference between pairs of means is recalculated at each step. The number of groups that needs to be compared is reduced by one at each step

Duncan

Same as Newman Keuls but using the same Fcrit as sidak

Scheffé

More conservative, groups of different sample sizes, less conservative to assumptions of homogeneity and normality are not met. Uses a larger critical value than for the omnibus test (dfs) x (k-1)

Turkeys HSD

Groups of equal sample sizes, all comparisons represent simple differences between two means, uses the Q statistic.