I'm new to multilevel modeling and would be grateful for some feedback on a twolevel path model (manifest variables only) that I am trying to fit.
Our hypothesis is that y is explained by a predictor x both on the individual and the team level. To differentiate between these effects and to ensure there is no multicollinearity, we use a group-mean cenetered x in the within model (defined as a within variable) and aggregated team means of x (defined as a between variable) in the between model. Does this make sense?
Also, we would like to include a control variable c, which we believe to have an effect at the team level (besides some other control variables on the individual level). However, c was measured on the individual level. To control for the effect of c on team level, do I need to calculate team means for c and enter those team means in the between level? Mplus also gives a seemingly meaningful output if I stick the "raw" c in the between model without calculating means (but also without defining it as a between variable, as it obviously has variance within the teams) - but I am not sure what this means conceptually...? If I stick the "raw" c in the between model, should I also use it as a control variable in the within model?
thank you very much for your reply!! I'm glad to hear that the approach makes sense and I will enter the control variable c as a team aggreagted variable as suggested.
I do have two small follow-up questions though, maybe you can give me a hint on these too? Q1: If I enter the team aggregated control variable c in the between level, should I also enter the group mean centered control variable c at the within level? Q2: Just as a matter of interest, what does Mplus do conceptually when a user enters an individual level variable that is neither defined as within nor as between level variable as a predictor in a between level model? It does calculate estimates and does not show an error warning - but is this calculation meaningful in any way?