I did two a two way repeated measures ANOVA for a within-subjects cross over study to evaluate the effect of tDCS on HRV over time (40 minutes) in patients with Alcohol Use Disorder. With tDCS put on for 20 minutes between 10-30.
My within subject subjects factors are treatment (sham and real) and time (8 periods of 5 mins). I found only significance for time and not for treatment nor treatment*time.
I want to analyze between which moments in time their is a significant difference. But the pairwise comparisons don't show any significance between periods.
Can I do one way anova seperate for sham and real?
Do I use paired students t-test to evaluate difference in time points?
any other suggestions...
Thank you,
student
If you have any theoretical reason to run a one-way test for time twice, yes do so. Yet , do not use the paired t-test, but a Post Hoc test: the LSD test with adjusted p-values, e.g. with the BH ( Benjamini & Hochberg (1995)) method. In attachment an R script which illustrates how to do this. In this example Eye = your Treatment, and Strength = your Time. You will have to double the levels for Time: 8 instead of 4 levels (e.g., in the bar plot you will have to define 8 colours).
I would gree with Jos Feys above that you would need a reason for doing what you suggest. I might think of trying an ANCOVA if you have some reasonable covariates. I attached my fav resource on Ben-Hoch and Bonferroni procedures which you might find helpful. Best, D. Booth
- Badges
- Science topic