Interaction effect of group (IV) and questionnaire type (DV) - Possible with MANOVA?

#1
I already tried google and also the forum search - unfortunately I didn't find any good answer to my question:
I want to know whether my two groups differ on two subscales of the same questionnaire.
My hypothesis is that group 1 is worse on both subscales, but the difference to group 2 is bigger on subscale 1.


I have come up with the following solutions:
Solution 1: A multivariate ANOA with group as factor and the questionnaire subscales as dependent variables. In this case, I find a main effect of group. I also find that groups differ significantly on both subscales. But how can I compare the difference between the two differences? Is there a way to calculate an interaction effect? Or could I compare the effect sizes?

Solution 2: A repeated measures design with group as a between subject factor and questionnaire subscales as within subject factors. In this case, I can not only check the main effect of group, but also can investigate whether there is an interaction effect of group and questionnaire. But, as I am assuming that the questionnaire subscales measure different subconcepts, am I allowed to use this test?

Thank you very much for your help!
 
#2
Re: Interaction effect of group (IV) and questionnaire type (DV) - Possible with MANO

many people have read this posts, but unfortunately no reply yet - is there a reason for this? could I formulate the question more clearly or should I post in another forum?

maybe one more addition: usually, I always thought that a repeated measures ANOVA can only be used for one and the same time outcome measure. There are papers however that use within subject ANOVA to test an interaction effect between group and test type. in this article, researchers even argue that an interaction effect is the only way to prove that a group effect is present in one but not in the other condition:
http://www.nature.com/neuro/journal/v14/n9/full/nn.2886.html
 
#3
Re: Interaction effect of group (IV) and questionnaire type (DV) - Possible with MANO

There doesn't seem to be any interaction here, but you have 3 hypotheses, not 1.

Hypothesis 1: Group 1 is worse on subscale 1
Hypothesis 2: Group 1 is worse on subscale 2
Hypothesis 3: The difference between group 1 and group 2 is larger for subscale 1

I don't see why you need MANOVA at all; I don't see any repeated measures. Each of these hypotheses could be tested with a t-test (if assumptions are met) or some other test of means.
 
#4
Re: Interaction effect of group (IV) and questionnaire type (DV) - Possible with MANO

Dear Peter,

thank you for your reply!
I am not quite sure I understand how to test hypothesis 3 though.
To perform a t-test, i need a mean and some sort of variance. I see how I can calculate the mean difference between group 1 and 2 for each scale, but how can I calculate a standard deviation for the difference?
 
#5
Re: Interaction effect of group (IV) and questionnaire type (DV) - Possible with MANO

You have N people who took both subscales. You find a difference between the subscales for each person. Call this D. Now, D has a mean and sd. Compare D for subscale 1 to D for subscale 2.
 
#6
Re: Interaction effect of group (IV) and questionnaire type (DV) - Possible with MANO

Compare D for subscale 1 to D for subscale 2.
Did you mean: "Compare D for group 1 to D for group 2"?

Because if d is the measure of difference between two subscales, then I only have one d per person.

If I compare D between group 1 and 2, I can only say whether the difference between the two subscales is bigger for one group compared to the other, but not if the difference between the two groups is bigger on one scale compared to the other. This does make a difference because if I only know that group 2 differs more on the two subscales I still do not know on which one of the two subscales they are more similar to group 1.