2-way repeated measures ANOVA

JorgeTT

Hi everyone.

Running a 2-way repeated measures ANOVA, 2 groups, measurements pre and post.

Time * group tells me if the changes over time are different between the groups.

However, is JASP unable to tell me what is the amount of that difference? (only can tell me the partial eta square).

Is there an option to inform this?

Or is it more an actual limitation of the test itself? In order to know that I would have to run ANOVA or ANCOVA or t-test of the deltas, for example?

Thanks,

JT

EJ

Hi JT,

When you say you want the "amount", I assume you'd like to standardized measure of effect size? -- because you could of course just report the cell means. I am generally a bit cautious about interactions that do not cross, because they are not robust to monotonic transformations of the dependent variable. Also, I generally think it is a good idea to decompose an interaction to get a better idea of what is going on.

That said, I came across this post by Daniel Lakens where he calls attention to Cohen's f (http://daniellakens.blogspot.com/2020/03/effect-sizes-and-power-for-interactions.html) I just glanced at this though, and it wasn't clear to me how this measure is an effect size specifically for interactions. Maybe the Maxwell book has more background information.

Cheers,

E.J.

JorgeTT

Thanks EJ. I was more talking about Mean differences.

Suppose it is about Kg. In a clinical trial we would be more interest in the difference between the treatments, rather than the cell means, that actually may add little value (?).

From eta square of the interaction, I can back transform to SMD too. That would give the SMD value of the deltas, right?

"Contrasts" also are not able to give me the difference between the change in means?

I will check Lakens :)

Thanks