Lu Gram posted on Sunday, April 10, 2016 - 2:12 am
I'm trying to understand what the between-level regression of Y on X means when we simultaneously have a within-level regression of Y on X. E.g. in the following model:
ANALYSIS: TYPE = TWOLEVEL;
MODEL: %WITHIN% ylifesat ON mwlbal; !Within-level regression of the mediator and outcome
%BETWEEN% ylifesat ON mwlbal; !Between-level regression of the mediator and outcome mwlbal ON wctrygendereq;!Between-level regression of the IV and the mediator ylifesat ON wctrygendereq;!Between-level regression of the IV and the DV;
For me, both between and within-level regression reduce to adding a fixed effect to the regression model, so I don't understand how the model can be identifiable when you have both regressions in the model? But if I remove the between-level regression, Mplus adds a correlation instead.
When y and x are not put the on WITHIN or BETWEEN list, there is a latent variable decomposition of them into within and between parts which are used in the within and between parts of the model. See Example 9.1 for further information.