Interpreting Interactions in Mixed Model ANOVA with 3-level between-subject variable

[Kerwin Olfers]

Dear all, 

I’m sure this has been asked and answered before, but I just can’t seem to find a complete answer anywhere, so I’d greatly appreciate a nudge in the right direction!

I’m running a Mixed Model ANOVA with the following factors:

Within-subjects: Session(2), Task condition1 (2), Task condition2 (2)

Between-subjects: Group (3) 

Following the Discovering Statistics Using SPSS book, I’m able to interpret interaction of Session x Task conditions, using simple contrasts. 

However, I’m a bit at a loss of how to do this for Session x Group, and Session x Group x Task condition1 interactions, as there are no contrast specifications for Group.. I can see that the interactions is significant, I just don’t know between which groups the difference are significant. (As the book only shows the situation with a 2-level between-subjects variable)

What would be the appropriate course of action here? 

– Would it suffice to run the Mixed Model three times, just excluding 1 of the three groups each time? 

– Should I create difference scores, and then do a regular ANOVA?

– Is there a different, more suitable approach?….

Many thanks in advance!

cheers,

Kerwin

[Dave Collingridge]

Kerwin,

You may not have set up your variables properly. You need the following 9 variables in your data set where “s” represents session and “t” represents task condition.  

group
s1_t1.1_t2.1
s1_t1.1_t2.2
s1_t1.2_t2.1
s1_t1.2_t2.2
s2_t1.1_t2.1
s2_t1.1_t2.2
s2_t1.2_t2.1
s2_t1.2_t2.2

In SPSS go: analyze -> GLM -> repeated measures -> enter the following within subjects variables (number of levels shown in parentheses: session(2), task1(2), task2(2) -> enter a name for your outcome variable -> click define and move group into between subjects box and all other variables into the within subjects box (there should be slots for all 8 variables -> in contrasts change all contrasts to simple. 

This will produce all main effects, 2-way, and 3-way interactions in the output file.

[Kerwin Olfers]

Hi Dave, 

Thank you for your reply! 

I think I’ve set up the variables correctly, just to clarify, the within-group contrast work fine, I get all the interactions. However for the within x between interaction, say for the interaction Session x Group, I would get as a contrast: 

Session  (level 1 – level 2) x Group  

That is, it tells me there is a difference between session 1 and session 2 on the scores of the Different groups, but it does not give contrasts nor post-hoc comparisons to see between which of the 3 Groups the difference is significant. 

Thank you again,

Kerwin

[Author]

Posts