I conducted a 2x2x2 mixed-design ANOVA that revealed a predicted three-way interaction. I'm now a little unsure about how to proceed.
Post-hoc tests aren't available as there are only two levels for each IV.
Howell & Lacroix (2012) suggest simple effects tests using SPSS Syntax. Is this still the preferred method? I was considering splitting the file on one of the IVs, giving me two levels to run 2x2 ANOVAs on, but I'm not entirely sure that's acceptable.
I was wondering what your opinions were on the matter and whether you suggest simple effects or not.
Hello Ross,
If there indeed is a 3-way interaction, that implies that the two-way interactions are inconsistent across levels of the third dimension. I'd start by plotting cell means (you can certainly do this within spss via dialog box without any fuss or syntax). Eliciting all possible simple effects does require a small amount of syntax, however (and this would be far preferable to running two-way anovas).
Here's a good starting link for the unianova syntax for simple effects:
https://www.ibm.com/support/pages/significant-interaction-anova-how-obtain-simple-effects-test
Good luck with your work.
A two-way interaction is a difference in the simple main effects of one of the variables as you change levels of the other variable (if you look at just the graph of the women in Fig- ure 22.1, each line is a simple main effect).
I think simple main effects would be the most natural progression to your original analysis.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA