LGM with 2 sets of Time Varying Covar...
Message/Author
 ZHANG Liang posted on Saturday, January 05, 2013 - 7:19 am
Dear Muthen&Muthen:

As is noted on Page130, User's guide V7, a set of Time Varying Covariates (a21-a24) could be included in LGM modeling. Here, my question is, whether can I add an extra set of TVCs, e.g., b1-b4, into the model? Could I write the syntax like below?

MODEL: i s | y1-y4 AT a11-a14;
st | y1 ON a21;
st | y2 ON a22;
st | y3 ON a23;
st | y4 ON a24;

st | y1 ON b1;
st | y2 ON b2;
st | y3 ON b3;
st | y4 ON b4;

If my opinion is right, should I set all the TVCs correlated with each other?
 Bengt O. Muthen posted on Saturday, January 05, 2013 - 4:20 pm
I would think you want a different random effect for the b tvc's than the a tvc's. So you would say:

st | y1 ON a21;
st | y2 ON a22;
st | y3 ON a23;
st | y4 ON a24;

stb | y1 ON b1;
stb | y2 ON b2;
stb | y3 ON b3;
stb | y4 ON b4;

The tvc's are covariates (on the right-hand side of ON) and are therefore automatically correlated.
 ZHANG Liang posted on Sunday, January 06, 2013 - 12:24 am
Thank you for the advice, sir. Today I'm wondering whether I can discard the growth factor S in the LGM model. Here is the syntax for the MODEL part:

i by y1-y4@1;

st | y1 ON a21;
st | y2 ON a22;
st | y3 ON a23;
st | y4 ON a24;

i st on x;
i with st;

For example, y1-y4 are children's depression level at four time points, a21-a22 are the frequency of being bullied by peers at four time points. And if the intercept of st is estimated as significant, could I interpret the result like this: "Depression is dynamically associated with being bullied overtime", or "Depression and being bullied 'tracked' together across the course of study"?
 Bengt O. Muthen posted on Sunday, January 06, 2013 - 5:32 pm
If you don't have a trend over time, you don't need s.

I would simply say "Depression at a certain time point is influenced by bullying at the same time point".
Post:
This is a private posting area. Only registered users and moderators may post messages here.