I have the following study design:
- 1 between-subjects factor (3 levels; group A, B and C): Three distinct subject groups
- 1 within-subjects factor (3 levels): Response accuracy in a computer task
I analyzed the data with a mixed design repeated measures ANOVA approach. The analysis produced significant main effects for both factors (between and within), but no significant interaction effect.
When I look at the plotted data for the three groups, it is evident that the within-subjects main effect is driven by the data of group A. In this group, I see a linear increase across the levels of the within-subjects factor. Groups B and C do not show this increase.
This descriptive observation is also evident when I run (exloratory) separate one-way repeated measures ANOVAs for each group: There is a significant effect of the within factor for group A, but not for group B and C.
This is a potentially interesting observation. However, the lack of an interaction effect in the initial ANOVA prevents me from performing (i.e. reporting) the subsequent group-specific analysis.
How can I proceed with this? Would it be okay to run a second ANOVA with only two groups, report the potential interaction effect and perform post-hoc tests? Would I have to adjust the p-value?
Thank you for any help / suggestions!