t-tests after three-way ANOVA?

#1
Hi all,

I've just run a three-way ANOVA on my data (3 IVs, 1 DV). I've got some main effects, as well as a three-way interaction effect. Am I supposed to run t-tests after the ANOVA in order to work out where the significant interaction effect is stemming from?

Thanks
 
#2
Hi all,

I've just run a three-way ANOVA on my data (3 IVs, 1 DV). I've got some main effects, as well as a three-way interaction effect. Am I supposed to run t-tests after the ANOVA in order to work out where the significant interaction effect is stemming from?

Thanks

First, it's important that you include all 3 of the 2-way interactions that are "building blocks" of the 3-way. Second, doesn't your software break things down to give you parameters associated with each level of each interaction? For example SPSS will do so as long as you specify "/print parameters" or choose the corresponding menu option.
 
#3
Thanks for the reply.

So are the t values provided by parameter estimates the same as those that would be obtained from running an independent samples t-test? (that's what it looks like)
 
#4
Not exactly. the Independent Samples T-Test is a bivariate test, whereas your 3-predictor model incorporates control for other predictors. In fact, the 3-way interaction results control for all 3 main effects as well as the 3 2-way interactions. hth