Hierarchical LV Structural Modeling +... PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Eva posted on Saturday, May 26, 2012 - 1:16 pm
Hi. I have clustered data that I would like to analyze. At the individual level, I have some measured binary covariates and 4 factors (measured by 12 variables), with a binary outcome. At the group level, I have 2 factors (one that is the same as a level-1 factor, one that is made up of individual measures + group-level only measures), and other continuous measured covariates.

My theory concerns this level-1 binary outcome as predicted by both level-1 and level-2 variables, and then hoping to see if cross-level mediated relationships (e.g., 2-1-1, 2-2-1-1) exist. My model also involves multiple endogenous variables (so various 2-2-2 or 1-1-1 relationships as well).

Does this kind of problem sound like something Mplus can handle? One of the biggest mysteries to me in SEM is how the intercepts fit in the regression analyses. I am more familiar with HLM in which I would predict the level-1 intercept using level-2 predictors -- does Mplus accommodate intercept-as-outcome modeling, or is this framework not one in which I should be conceptualizing when doing the HLVSM?
 Linda K. Muthen posted on Monday, May 28, 2012 - 8:38 am
The intercept is a dependent variable on the between level just as it is in HLM. See the examples in Chapter 9 of the user's guide where random intercepts are shown as circles on the between level.
 Eva posted on Monday, June 04, 2012 - 9:19 pm
Thank you. Another question: How would I model an intercept of a within latent factor (not class) as a dependent variable at the between level?
 Linda K. Muthen posted on Tuesday, June 05, 2012 - 11:45 am
Random intercepts can be used as dependent or independent variables.
 Eva posted on Tuesday, June 05, 2012 - 2:54 pm
Thank you, but how exactly do I specify that in my syntax? Do I just take a factor that I specify in the WITHIN part of my model and use it as a dependent variable in my BETWEEN equation for it to be modeled as the intercept-as-outcome equation?
 Linda K. Muthen posted on Wednesday, June 06, 2012 - 10:57 am
A factor does not have a random intercept on between. Observed variables on within have a random intercept on between. You would need to declare a factor on between. See the examples in Chapter 9 of the user's guide.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: