Hi,
I have three independent variables directly measured from the students, on different type of perceptions. The dependent variable is teacher's assessment of students' class participation. The basic idea was that in addition to the subjective measurement of perceived participation, we ask teachers to rate students' class participation as well. The relationship between IVs and DV to be examined using structural equation modelling (SEM).
There were more than one teachers involved in the rating in such a way that one student can only be rated by one teacher, however, one teacher can rate multiple students.
My question is should I consider multi-level analysis? And, how can I control the DV for the independence of observer's ratings?
Any pointer/guidance to example study will be highly appreciated.
Note: I prefer to run SEM in SmartPLS, however, I believe multi-level SEM cannot be run in SmartPLS (right?). I can use AMOS as an alternative tool. I don't have access to MPlus.
Thanks in advance.
Best,
Ali
I consider that you can make two models one of the qualification of the teachers and another of the students, also I have seen studies that use both in the same transverse SEM analysis.
Thank you, @Cristian Ramos-Vera. I wonder if you could provide links/references/titles of mentioned studies? It will be helpful. Thank you.
I sounds like you need to use multi-level modeling (hierarchical linear modeling) due to the lack of independence of your observations -- students nested within teachers.
Thank you, David. Indeed it us a case of hierarchical linear modelling.
I have a supplementary question:
If there are two IVs, 3 mediators, and moderator between IVs and mediators (all from students) and 2 DVs (from teachers). How can I proceed with such complex model while hierarchical modelling is required?
Should I run a step-wise regression model?
Be careful not to confuse "hierarchical linear modeling" (which requires specialized forms of regression) with "hierarchical regression" (which involves the order in which independent variables are entered in ordinary regression.
I am not aware of any way to apply stepwise regression to multi-level modeling.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling