I have an FTIR spectral data set that I want to analyze with MANOVA in JMP Pro. The statistical model for the experiment contains two fixed effects and one random effect. MANOVA doesn’t support random effects. The experiment measured spectra of the same subjects over a period of 12 weeks. However, only the last 3 weeks are averaged and included in the analysis. So, I guess it is not considered as a repeated measure anymore.
Will MANOVA be appropriate for the analysis of this dataset in this case?
Is there a different procedure to do a multivariate analysis of variance on this data set that supports random effects?
Or should I use MANOVA and exclude the random effect? Is the random effect relevant in the averaged data?
Or is there a workaround to include the random effect in the MANOVA analysis in JMP Pro in this case?
Thank you
I don't think you want MANOVA. I think you want a linear mixed model. Here is a link to some course notes prepared by Prof Marie Davidian, a leading expert in the field:
https://www.stat.ncsu.edu/people/davidian/courses/st790/notes.html
I believe this material will be of help to you. Please look it over. Best wishes, David Booth
Thank you for your reply David, I really appreciate it and I will look them over. Anyway, I have already tried Proc Mixed from SAS; however, it only support one dependent variable. It doesn't take more. I already looped over all the variables that I need to test. I was looking for an alternative that can analyze multiple responses in one procedure.
The model, as mentioned in the question by Mazen Bahadi, Is neither a purely fixed effect model nor a purely random effect mode but a mixed effect model as I have understood. Accordingly MANOVA, as I think, will not be appropriate. However, it can be thought of an attempt on applying ANOVA for multi-way classified data in the instant problem.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS