Second order growth model
Message/Author
 Jon Heron posted on Wednesday, July 25, 2012 - 4:33 am
Hi Bengt/Linda

I am attempting to fit second order growth models to two processes in parallel with directional associations between the various growth factors.

Say I have first order factors:- b1 b2 b3 b4 b5 b6, a1 a2 a3 a4 a5 a6

This model is stretching both me and my PC (mainly me) so I am currently contemplating a two-stage approach where I fit a simpler model that covaries all my first order factors (ai,bi). I then read the tech4 output back into Mplus and fit the parallel growth model to these vars/covs/means.

I have a feeling this approach may have some shortcomings but is still useful/valid for informing a full one-stage model -e.g. in terms of my chosen polynomials and res-variance constraints.

Does this sound reasonable or am I missing something?

cheers, Jon
 Linda K. Muthen posted on Wednesday, July 25, 2012 - 10:08 am
You can do that. Your standard errors will not be ML and may not be correct.
 Jon Heron posted on Wednesday, July 25, 2012 - 10:18 am
thanks Linda
I need to do something.

bw, Jon
 lamjas posted on Monday, January 21, 2013 - 12:36 am
Can I do a LGM on a second-order factor?

For example (as an unconditional model):

Variable:
Names are
ay11-ay15 !Measured at T1
ay21-ay25 !Measured at T2
ay31-ay35 !Measured at T3

by11-by15 !Measured at T1
by21-by25 !Measured at T2
by31-by35 !Measured at T3

cy11-cy15 !Measured at T1
cy21-cy25 !Measured at T2
cy31-cy35; !Measured at T3

Model:
fa1 by ay11-ay15;
fa2 by ay21-ay25;
fa3 by ay31-ay35;

fb1 by by11-by15;
fb2 by by21-by25;
fb3 by by31-by35;

fc1 by cy11-cy15;
fc2 by cy21-cy25;
fc3 by cy31-cy35;

fabc1 by fa1 fb1 fc1;
fabc2 by fa2 fb2 fc2;
fabc3 by fa3 fb3 fc3;

s i | fabc1@0 fabc2@1 fabc3@2;
[fabc@0 fabc2@0 fabc3@0];
 Bengt O. Muthen posted on Monday, January 21, 2013 - 8:12 pm
Yes. It will be identified if you hold observed item intercepts equal across time and fix 1st-order factor intercept at zero.
 Mahdi posted on Monday, June 02, 2014 - 12:11 pm
Dear Dr.Muthen;
You explained about "Multiple indicator linear growth model for continuous Outcomes" and "Two-level growth model for a continuous outcome (three-level analysis)" in example 6.14 and 9.12, respectively in your excellent user's guide. Can we run multilevel multiple-indicator linear growth model in Mplus? Please introduce me a reference about that??
Thanks a lot.
Mahdi
 Linda K. Muthen posted on Tuesday, June 03, 2014 - 10:53 am
We don't have a reference for that. You would need to put it together from the two examples.
 Mahdi posted on Tuesday, June 03, 2014 - 11:09 am
Hi,
Thanks, for this answer. In my model all indicators are continuous, Is ESTIMATOR=WLSM (or WLSMV) correct?
 Linda K. Muthen posted on Tuesday, June 03, 2014 - 1:39 pm
Neither. You should use MLR.
 Mahdi posted on Tuesday, June 03, 2014 - 2:00 pm
But when I used MLR method, the model did not converge and this warning is appeared:

WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE
DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A
LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT
VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES.

THE H1 MODEL ESTIMATION DID NOT CONVERGE. SAMPLE STATISTICS COULD NOT
BE COMPUTED. INCREASE THE NUMBER OF H1ITERATIONS.

What's the reason?
 Linda K. Muthen posted on Tuesday, June 03, 2014 - 2:53 pm