
Message/Author 


Hi. I'd like to estimate a twolevel model (people clustered with supervisors). I have specified a unidimensional model (for ordered categorical items) on the within level. I am aware of how to let loadings vary. However, how do I specify a model that lets the factor mean vary across clusters? Sorry if I'm being dense. Here's a code snippet: %within% g1 by sf1* sf2 sf3 sf4 sf5 sf6 sf7 sf8 sf9 sf10 sf11 sf12 sf13 sf14; g1@1; %between% How do I let g1 vary across clusters (without getting an error)? To be clear, I don't hypothesize a cluster level factor. Rather, I want to estimate how much the factor varies within and across clusters. For example, suppose y1 us a continuous variable. The following code would result in an estimate for the mean and the within and between level variance: %within% y1; %between% y1; How can one do the same with the factor g1? Thanks! 


y1 refer to the variance not the mean. In a crosssectional study, the mean of a latent variable is fixed at zero. 

Back to top 

