The Games-Howell post hoc tests and confidence intervals do provide protection for the set of pairwise comparisons among means in a one-way between-subjects design, though overall protection at the stated level is guaranteed only in infinite sample sizes. In finite sample sizes, the tests and intervals may be slightly liberal, but overall seem to provide good protection with a little bit more power than alternatives. This is achieved by using what might be considered some modifications to the Tukey-Kramer approach. That is, the studentized range distribution is used to obtain critical values, but the degrees of freedom are based on Welch's approximation, and instead of using the pooled mean square error in forming the standard error of the difference, separate variance estimates are used.

I am love the game and reading this informative blog thanks sharing this and I am create a new game app more help check it website movie box