Finding the INTER class correlation for students assessment of classroom activities

[Maleika Krüger]

Hi everyone,

I have a dataset with students within classes. I calculated the intra class coefficient already (as a part of the multilevel analysis).

But now I also want to know how similar the answers within the classes are.
Backround: The students were asked to rate the frequencies for different classroom activities in the last few month (e.g. reading a book, watching a movie etc). I want to know if the students in one class experienced the lessons in the same way/remember it the same way. Obviously a great discrepancy would be a bad thing for me, since it would mean, that students in the same class perceived the lessons very differently.

I can't come up with a procedure to get what I want. Do I need to calculate a correlation seperately for each student with every other student in his/her class? Do I need to reshape the dataset to show every class in one row, almost treating it like a inter-rater reliablity test where every students is one 'rater' rating the same issue?
Any suggestions would be greatly appreciated

[Clyde Schechter]

It sounds like you are trying to quantify the extent to which students in the same class provide similar ratings. The intra-class correlation is one measure of that, though, in effect, it does it by comparing the variation within classes to the variation between classes. So, even if students in the same class rate the activities very similarly, if students in different classes also rate the activities very similarly, then the intraclass correlation will be low. It may be that you just want to look at the within-class standard deviation.

A fuller description of your data, your research question, and the model you estimated might facilitate a more specific answer to your question.