Two-way ANOVA Levene's test - Statistica General Discussion - Statistica - Dell Community

# Two-way ANOVA Levene's test

#### Two-way ANOVA Levene's test

Hi,

I am currently working with Statistica 12 and have the following problem: I have two binary factors X1 and X2 and one continuous response variable Y and I try to perform a two-way ANOVA analysis without interaction between the factors. When I want to check the assumption of constant variance, I go to the 'Assumptions' tab to get the result of Levene's test. But Statistica doesn't let me select the X1*X2 effect! I am only allowed to select either a X1 effect or a X2 effect.

So how do I check whether the variances are equal in the four groups using Levene's test? Or is it not nessecary to have equal variances in the four groups in a two-way ANOVA without interaction?

Tony

All Replies
• How did you specify the design?  Are you choosing Factorial ANOVA?  If so, it should automatically include the interaction.  That would be easiest.

If you are using GLM, without choosing Factorial ANOVA, you will need to click the button for between effects and specify the main effects and interaction to test.  Then you can test the interaction for HOV.

• Thank you for your quick response. I chose 'Main Effects ANOVA' because I didn't want to include an interaction term anyway. I see now that I can indeed test HOV in the four groups using 'Factorial ANOVA'. But why is this test not included in the Main Effects design?

• Because in main effects you won't test for an interaction.  So if you have one DV (y) and two factors (x1 and x2), and you have a main effects design only, you won't test for an interaction.

• But Levene's test is not there to test the interaction. It's to test the assumption that the variance is equal in all four (2x2) groups. And that assumption should hold, even in a main effect design, right? So I still find it strange that there is no option to test the following hypothesis:

H0:  σ11=σ12=σ21=σ22