Hi, I am pretty new to Mplus and I have a question on centering and clustering.
I am working with a large sample of 32 schools, around 280 classes, and roughly 6000 pupils.
We developed a new task that is assessed in the regular classroom (WM), so we are expecting classroom effects on performance. We also expect that the predictive value of the task (WM) for academic achievement (AA) may have classroom effects. I do not have any classroom variables available at the moment, and I just want to use class as a cluster.
In a basic analyses, I have found an ICC of .376 which is large. Now I am wondering whether age differences between classes might excerbate this value,since whe have classes varying from grade 1 thru grade 6. Is this indeed a problem? and if so, what would be the best way to solve it?
I have centered task performance on the grand mean, but perhaps I should be using a different mean (age mean, grade mean, class mean?)? Or is there a different way around this? The fact that we have classes with mixed grades (e.g., grade 3/4) as well as homogeneous classes as well, might make this more complicated.