I have a repeated-measures crossover study design with 3 treatment conditions, all measured across 7 time points.
I would like to control for baseline for each treatment. As I understand, a normal ANCOVA design is not able to do this. Which test would I use in SPSS?
You can model change from baseline and leave baseline variable as a covariate in the model if the three treatment conditions are randomly assigned such that you would expect the baseline of three groups are same. You can use mixed model in SPSS. In Fitzmaurice Garrett's book "Applied Longitudinal Analysis", section 5.6, you can find the discussion about this topic with great details.
Thank you for your answer.
Could you just explain why I would use a mixed model? Where does my 'between-factors' come from?
You can check out the attached article for some details. This is about SAS but the idea is same. For your study, you need to add time effect in the model.
Terun, your treatment conditions is a between subjects effect.
A mixed model does not necessarily require a between subjects effect on the other hand. A mixed model is not a mixed Anova, it is more general. It is called mixed because it may contain random and fixed effects.
@Mehmet
If the concept of within-between subjects does not apply to the mixed model, then surely I assume that 'treatment' should be my fixed effect and 'time' my random effect. Is this correct?
Why would my treatments then be a 'between-subjects effect'?
How many sequences of the treatments were incorporated in the crossover? Is your outcome continous or binary?
In general, with a continuous variable, mixed-effects linear model would be needed to account for the repeated measurements that yield period (time), sequence, and carryover effects and to model intra-individual and inter-individual variability. With a binary outcome, you would need GEE.
To be clear, do you mean that each treatment in Sequence 1 had 7 repeated measures for the outcome prior to introduction of the next treatment in Sequence 1, and the same for Sequence 2, 3, etc.?
@Robert
I have 3 treatments. All participants underwent the 3 treatments in a random order.
For each treatment I took measurements 7 times.
@Angela
All participants underwent all treatments hence I don't understand why this would be a between effects, as the same person has done all 3 treatments.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis