Hello Ineke,

In the anova framework, continuous IVs are entered as covariates. Be sure to specify that you want interaction/s involving such terms as part of your model.

You may find it easier to express the analysis as a regression model (same inherent linear model), so that you can include all the relevant interaction terms explicitly (Treat*Mod 1, Treat*Mod2), and decide for yourself whether mean-centering is desirable or not. The treatment and categorical moderator variables would just be dummy coded values.

Finally, Andrew Hayes' Process add-in will do much of the heavy lifting for you (see: http://processmacro.org/index.html); versions are available for spss and sas (and a version for R is coming, if not currently included).

Good luck with your work.

Very true, David.

Hello Ineke,

this can be easily analysed in a regression or SEM framework with a dummy (Treatment / control) x moderator product term.

If your moderators are measured with multiple indicators and adher to a common factor, it has severe advantages regarding power if you switch to a SEM framework with latent variables (although I would start with a quick/dirty regression model :)

Best,

holger

Hi all,

thank you so much for your replies!

I understand your comments. However, I want test the two moderators at once. I will explain more clearly.

IV = treatment/control (email sent or not)

MOD 1 = rational/emotional content

MOD 2 = time being inactive. (which is from 1 month to 1 year)

So, I also want to test rational x time being inactive for the treatment. And emotional x time inactive.

Is that also possible with a regression?

Thank you a lot!

Hello Ineke,

yes. In that case you would add two product terms and if you are really bold :) a three-way product.

There is nothing in using dummies or conducting an experiment that conflicts with the overall general linear model. In this regard, ANOVA (that nobody knows outside psych) and regression are just two different version of the same thing.

Best,

Holger