exchangers!
I am struggling with my data analysis for an article I'm writing. I have to following experimental set-up.
- Instruction (A/B)
- Group (Alcohol Dependent / healthy Control) – between subjects facotr
- Medication (C/D) – (between subject Factor)
- Session (Baseline / Medication) – Within Subject factor
All subjects performed a certain taks, related to evaluating emotional pictures in a certain way (dependent on instruction A or B). Medication was only administered at the medication (2nd) session. Analyses are performed with SPSS v20.
I would like to perform separate analyses for Instruction A and B, but the 4-way interaction is not significant (when putting everything in a 4-way repeated measures anova). This mainly says that the effect of Medication*Session does not differ between groups or instruction. If so, can I perform separate analyses for instruction? Instruction A and B are very significantly different (assessed with a t-test, for the baseline session so medication has no efffect on this). I would want to argue that since, these processes are so different it is better to assess the effect of medication, group and session separately for instruction A and B. The problem is that the non-significant 4-way interaction does not support this claim.
Many thanks in advance!
Best Answer
Four way interactions are a mess to interpret.
You can do separate analyses regardless of whether the 4 way interaction is significant or not.
You say you are using repeated measures ANOVA - have you checked all the assumptions? They are very often violated and that might affect the results.