I am trying to run a micro-macro multilevel model, as outline in Croon and Van Veldhoven (2007). I am predicting a school-level fidelity variable (FID) from individual teacher level scores on a school climate measure (ILS). I would like to model ILS as a latent variable with 7 subscales (indicators). To the best of my understanding, I need to compute adjusted means of each of the ILS indicators in my level-1, within-level model, and then use these adjusted means to create a level-2 between-level latent factor of ILS which I then use to predict my level-2 fidelity score. Is this correct? If so, do you have an example of Mplus syntax that I could apply to this model?
You can specify an ILS factor on Within (ILSW) and an ILS factor on Between (ILSB) and hold their factor loadings equal across the two levels. Don't fix any loading at 1 but instead fix the variance of ILSW at 1 and estimate freely the variance of ILSB. Your Between-level FID variable can then be regressed on ILSB on the Between level.
To clarify, I should be creating an ILSB factor in my between-level model? I'm getting an error that I can't use within-level variables (which all of the ILS indicators are) in the Between section of the model.
"Within-level variables cannot be used on the between level."
Here's my syntax if it would be helpful. Thank you so much! Larissa
USEVARIABLES ARE FIDper ilspro ilskno ilssup ilsper ilscom ilsvis ilsava;
MISSING = ALL (-99);
WITHIN = ilspro ilskno ilssup ilsper ilscom ilsvis ilsava;