|
|
 |
| Nonlinear constraints with a count va... |
 
|
| Message/Author |
|
|
|
|
Drs. Muthen,
I am trying to use Mplus to run a moderated nonlinear factor analysis (MNLFA), as described in Bauer and Hussong (2009), for which I have included sample code below. In this model, the measurement parameters relating an indicator (y1 below) to a latent factor (eta below) are affected by a covariate (x below). Additionally, the variance of the latent variable is affected by this covariate.
In the past, I have been able to run this exact model with continuous, ordinal, and binary outcomes. However, in this particular case, y1 is a count variable, and the model will not estimate. (In particular, I get an error message which says, "FATAL ERROR: THIS MODEL IS NOT AVAILABLE.") I've tried a bunch of workarounds, such as using the XWITH parameterization, but have had no luck.
Do you have any ideas or suggestions for how I might be able to fit this model using Mplus software?
Thank you so much in advance!
===============
SAMPLE CODE:
Variables:
names = x y1 y2 y3 y4;
count = y1;
Model:
eta by y1* y2* y3* y4*;
eta by y1*(L1);
eta*(V);
[eta@0];
y1 ON x;
Model constraints:
new (L10 L11 V1);
L1 = L10 + L11*x;
V = exp(V1*x); |
|
|
|
|
| The Constraint= option is not yet developed for Counts. |
|
|
| Back to top |
|
|
