I am searching for a software package which is able to analyze a data set containing a cross-classified structure in the second level and third level.
We got 5 levels:
level 1: cases (N= 15632),
level 2a: students (N=801), level 2b: tasks(N=34),
level 3a: classes(N=30), level3b: topics (N=3),
level 4: teachers (N=23),
level 5:schools (N=19).
Is it possible to perform structural equation modeling using all levels of this data set? By now, we devided the anlyses into two steps: For the cross-classified multilevel-part we used MLwiN and for the second SEM part we used MPlus. Now we are searching for a software to combine these two parts. Any suggestions?
Thank you very much!
Hi Leonhard!
Thank you very much for your anser.
I added the sample size to my initial post.
Best,
Matthias
MLwin as you say can handle the 5 level structure - what are the bits of SEM that you need to do?
We got two latent variables on student level, which predict 3 further latent variables on the same level. Until now we used an ESEM-measurement/SEM approach to map this part using Mplus.
Hi Matthias,
I undertook an SEM and a multilevel analysis with AMOS and MLwiN respectively. Given your data structure I suppose an SEM would be futile. I can recommend MLwiN, as it can handle highly unbalanced samples. Whether a structure is cross-classified is not always clear upfront. MLwiN can help you out with this with its built-in MCMC procedure: try (non) cross-classified models and stick (as a rule) to the models with the lowest DIC statistic. Looking at your structure I wonder whether it will be cross-classified indeed. Do the pupils always stick together if they go to a higher level? That would be rather exceptional, I think. So I would also try nested models.In MLwiN that is quite easily done, by unchecking just one tick-box.
Kind regards, Paul Melessen
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling