Hi everyone,
I am studying the effects of odor presentation during sleep and wakefulness on psychotherapy outcomes. My experimental design includes: * A between-subject factor: Group (sleep or wake) * Two within-subject factors: Odor (congruent or incongruent) and Time of measurement (pre, post 1, post 2). For all my psychotherapy measures, I conducted a 2x2x3 ANOVA (Group x Odor x Time). However, I found that for some variables, there are significant/marginally significant differences between the wake and sleep groups at baseline (please see the attached figure for an example). To address these baseline differences and run an ANCOVA, I need to identify an appropriate covariate that works for my between-subject factor(sleep-wake). Here are my questions: What can I use as a covariate in this context? Is it correct to think that I need to perform ANOVAs separately for each sleep and wake group, and that I cannot include a covariate in my 2x2x3 ANOVA? I hope I was clear enough. I will be looking forward to your response and I thank you so much in advance!
I think a mixed-design analysis of variance (ANOVA) is more appropriate in this case.
A mixed-design ANOVA allows you to assess the main effects of each factor (Group, Odor, Time of measurement) as well as any interactions between factors.
Specifically, you can examine:
The main effect of Group to assess overall differences between the sleep and wake groups.
The main effect of Odor to assess the effect of odor presentation.
The main effect of Time of measurement to assess changes over time.
Interactions between factors (e.g., Group × Odor, Group × Time, Odor × Time) to explore whether the effects of one factor depend on the levels of another factor.
ANCOVA will not correct per-exisiting group differences. Indeed, it is likely to produce a very misleading result. The purpose of ANCOVA is to increase the precision of an analysis by reducing the magnitude of the error term based on its relationship with a non-experimental variable, not adjusting per-existing group differences. My advice is to run multiple ANOVAs at different levels of what you planned to use as the covariate. There are several papers in my publication list about this.
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity