ezANOVA model specification: no interaction

[Andreas Busjahn]

How can I request main effects without interaction (or just interaction)? The 'normal' way depVar~a+b (or depVar~a*b-a-b) can not be specified using the parameter of the ezANOVA function.

Tnx in advance!

[Mike Lawrence]

This isn't possible with ezANOVA. So I understand your intent more clearly: if you could do this, what would you expect as output and/or what would you want to do with the resulting models?

[Henrik Singmann]

Hi Andreas,

you might want to check the aov.car() function from my afex package (on CRAN) which should give you what you want.
But note that loading afex it changes the default contrasts.

Cheers,
Henrik

2012/11/21 Andreas Busjahn <herr...@googlemail.com>

[Stefan Th. Gries]

>> How can I request main effects without interaction (or just interaction)?

> This isn't possible with ezANOVA. So I understand your intent more clearly: if you could do this, what would you expect as output and/or what would you want to do with the resulting models?

The idea could be to see what the results are when an insignificant interaction between two predictors has been removed (in a backwards model selection process).

[Thom]

In general you don't want/need to do this in classical ANOVA because the factors are orthogonal and hence the interaction SS is distinct from the main effects. If the ANOVA has non-orthogonal factors then it becomes more complex (e.g., how to partition SS etc.) - though you shouldn't in general remove the interactions if there are a priori reasons for expecting them to be important in the model.