I have a model in which two groups are measured on one DV at pre- post- and 3 months following an intervention (the control group did not receive the intervention). I have an interaction effect, which I believe implies that the combination of group and time together influence the DV, and yet the same SPSS ANOVA output shows no main effects. How should I interpret the data? Is there an easy way to do further analyses with SPSS?
$\begingroup$
$\endgroup$
Add a comment
|
3 Answers
$\begingroup$
$\endgroup$
Here is an example where there's really no interaction. That means that the effect of factor B is additive and is the same at each level of factor A (the difference between the blue point and the red point is the same across all levels of factor A).
The lines joining the means provide the so-called interaction plot. If there is an interaction then we should not be interested in the main effects (but be careful about the meaning of "no interaction").
answered Jun 13, 2013 at 9:54
$\begingroup$
$\endgroup$
The interaction means that across time the effect of the intervention is changing. That's generally what you want to find. It's possible you don't need anymore tests. If the effect of treatment is small or reversed in the pre and then larger in the post and 3 months then your interaction has shown this pattern to be significant. There's no further test needed.
Many more details, like the pattern of data in a little table, would help others assist in interpretation.
$\begingroup$
$\endgroup$
I always recommend plotting the interaction in an interaction plot for interpretation. For instance, you could plot means and error bars (e.g., means +/- one standard deviation) over participants for the three time points, separately for the intervention and control groups. Chances are that the relevant interaction will hit you right between the eyes.
After that, you can do post-hoc tests. These are slightly more complicated for mixed models than for non-mixed ones.
answered Jun 13, 2013 at 8:03