PDA

View Full Version : ANOVA Simple effects help

statss
12-07-2011, 04:11 PM
When finding a significant interaction effect in a factorial ANOVA and then doing a simple effects of all of your comparisons, do you need to divide alpha by the number of comparisons you are doing? Thanks

ledzep
12-07-2011, 04:13 PM
You mean when you are making pairwise comparisons?

statss
12-07-2011, 04:17 PM
If I am doing a 2 by 2 factorial ANOVA and I find an interaction. Are simple effects pairwise comparisons? If so then yes haa I am comparing each condition to each other. So if I have 4 different simple effects I'm doing would I need to dived alpha by 4?

ledzep
12-07-2011, 04:32 PM
By simple effects, you mean the main effects of each factor?

Type I error, which is the error of incorrectly rejecting the null hypothesis when it is true, increases with increasing number of test you perform. If you perform more comparisons, then by chance, some of them would incorrectly reject your null. Hence, adjustments should be made to the alpha when performing multiple comparisons.

There are many ways making the adjustments. Few popular ones are: Dunnett's, bonferroni, Tukey's, Sidak's ..

From a strict statistics point of view, I think you should adjust your alpha level...Here, you suggested dividing by n=4, which is in fact, Bonferroni adjustment.

statss
12-07-2011, 04:34 PM
Yes the main effects for each factor! Thank you.