

Different predictors at the two levels 

Message/Author 

Eva posted on Sunday, October 02, 2011  8:18 pm



Hello. I am proposing a model to predict a studentlevel (within) outcome that is binary. At the studentlevel, I have some observed demographic covariates and 2 latent variables predicting this outcome. At the schoollevel (between), I have some school variables (not indicators of latent variables) and then the same two latent variables as in the withinlevel, but one of them is going to be measured a little differently by using between variables + a within variable, whereas the other is measured by the same 3 within indicators as the latent variable in the withinlevel. In the betweenlevel, the observed school variables affect one of the 2 school latent variables, and both school latent variables will predict the withinlevel student outcome. In other words, one of the latent variables is hypothesized to mediate the effect between those observed school variables and the studentlevel outcome. In a lot of examples, I have seen people using the same exact same measurement and structural models at both levels, but mine will have some same but also some different variables in either level. Is there a similar example out there that I can reference? Thank you so much for your attention! 


I think it is likely that models on within and between differ particularly the number of factors. See the following paper that is available on the website: Muthén, B. & Asparouhov, T. (2011). Beyond multilevel regression modeling: Multilevel analysis in a general latent variable framework. In J. Hox & J.K. Roberts (eds), Handbook of Advanced Multilevel Analysis, pp. 1540. New York: Taylor and Francis. 

Back to top 

