I am conducting an experiment examining the effect of X on Y in and comparing this tratment against a control group. I would also like to examine how this effect is moderated by two other variables (say, A and B) which i am not manipulating, just measuring.now how would i proceed with the statistical analyses for examining treatment interaction effects? I see that there are various methods to do this,but i cant seem to figure out what suits my situation, especially since its not possible for me to carry out a factorial design.
Well was nervous about doing that since i read the assumptions of three-way anova and one of them was independence of observations.from this, i understood that i needed a seperate group for each of my variables and compare them against each other. Hence i would need a factorial design. Am i correct,or did i misunderstand? If i am,then its not possible for to carry out a factorial design
You can add A and B to a general linear model. This is like adding a block to the model (compare completely random block design (crbd)) or adding a covariate to the model (compare to analysis of covariance (ancova)). You can add interactions between A and X or B and X or other interactions as you wish.
Yes, you can include three way interactions, provided you have enough degrees of freedom (sufficient observations).
A mixed effects model may be appropriate. I don't have much advice here except to say it is a topic that requires some reading on. But also, if you aren't comportable with mixed models, keeping all factors as fixed effects and using an OLS general linear model usually gives results that are just as satisfactory.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis