I have a set of 6 related likert-type DVs, a within-subjects time variable (i.e. each DV was measured at 3 time points), and a categorical between-subjects IV (condition). Is there a way to include all 6 DVs in a single model, or should I conduct 6 separate regressions? Thanks very much
Start with this link: multivariate poisson regression - Bing
It should help you. Do all your DVs together if possiblle. Now you have a problem since these DVs are likert(ordinal). You want Multivariate ordinal logistic regression.
See: multivariate ordinal logistic regression - Bing I've attached a helpful set of notes on univariate ordinal logistic regression. Best, David Booth
Start with this link: multivariate poisson regression - Bing
It should help you. Do all your DVs together if possiblle. Now you have a problem since these DVs are likert(ordinal). You want Multivariate ordinal logistic regression.
See: multivariate ordinal logistic regression - Bing I've attached a helpful set of notes on univariate ordinal logistic regression. Best, David Booth
This is quite challenging - especially as not all software can handle this sort of model. It can be done in brms using an ordinal logistic regression set up as multivariate model. See:
https://psyarxiv.com/x8swp/
https://cran.r-project.org/web/packages/brms/vignettes/brms_multivariate.html
Thom Baguley David Eugene Booth I was thinking in line of "Ordinal Logistic Regression for repeated measure(3 time points)". How appropriate for his case, which will include all his DVs? And, could SPSS, SAS or R be more appropriate to handle these analysis. Thanks in advance.
Here is a practical application of a discrete outcome multivariate multilevel model- we used the MLwiN software
Article Investigating the Macro Determinants of Self-Rated Health an...
I do not get where the Poisson model of the original question fits in.
It can be done in R using the ordinal package or brms (and probably other packages). I'm sure there are options in SAS, but I've never used SAS. SPSS can handle these sort of data via a GEE and possibly via multilevel generalized linear models, but I've never tried the latter in SPSS.
The main issues are convergence for these sorts of models so MCMC estimation in, say, MLwiN or brms may be required.
Abiodun Christian Ibiloye I only have tried R. It worked well for me. Best wishes, David Boot
Hello Lauren. Your question reminds me of the classic article by Hubery & Morris (1989) on MANOVA vs multiple ANOVAs. In the spirit of that article, my question is this: Do you have one multivariate question or 6 univariate questions?
Article Multivariate Analysis Versus Multiple Univariate Analyses
As David Eugene Booth said, the choice is multivariate ordinal logistic regression for panel data.
However, I have another suggestion. Since you have just one IV, you can interchange your IV with your DVs and then perform multiple regression instead of multivariate regression.
Seyyed Amir Yasin Ahmadi I am not sure about your last post. It seems to me that would change the research question. Perhaps you can reply in a bit more detail. Best wishes, David Booth
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression