I am running a path analysis model examining multiple mediators in the link between poverty and school grades. All variables are measured at the individual level. However, individuals are taken from 16 different schools (with sample sizes ranging from 5 to 160 individuals per school). I'd like to control for school-level effects, but my understanding is that 16 clusters is too few to run a TWO-LEVEL or COMPLEX model. Is there another way to control for school level in Mplus?
You can either use 17 dummy covariates or turn to Bayesian analysis as described in the paper on our website:
Muthén, B. (2010). Bayesian analysis in Mplus: A brief introduction. Technical Report. Version 3. Click here to view Mplus inputs, data, and outputs used in this paper. download paper contact author show abstract
Tino Nsenene posted on Thursday, December 15, 2016 - 7:18 am
...interestingly, a new study claims that accurate inference is possible using REML estimation even if you have very few clusters; see