Hi all,
I face problems in specifying the correct Mplus input for running a multi-group comparison of a latent change score (LCS) model, including a PRE and POST measure with 3 indicators for each measurement point using robust ML with n=778 cases.
I follow Geiser's approach (see Geiser 2013: 145ff). So my model statement in Mplus for the entire sample looks like this (and works fine with good model fit and straightforward interpretation of estimates):
MODEL: PRE BY par11@1 !measurement model
par12 par13 (1-2); !incl. loading constraints for weak invariance
POST BY par21@1
par22 par23 (1-2);
DIFF BY par11@0; !define latent change/difference variable
[par11@0 par21@0]; !fixing intercepts to get latent means
[PRE DIFF]; !estimation of latent means
par11 with par21; !error-cov's across measurement occasions
par12 with par22;
par13 with par23;
![par11 par21] (3); !not needed because already fixed to zero
[par12 par22] (4); !intercept constraints for strong invariance
[par13 par23] (5);
POST ON PRE@1 DIFF@1; !perfect regression on post-measurment
POST@0; !variance of post-measure set to zero
For the multi-group analysis I used the same MODEL statement but
1. I added the lines
MODEL groupA: [PRE@0 DIFF@0]; !set latent means in group A to zero
MODEL groupB: [PRE DIFF]; !freely estimation of latent means in group B
2. and in the VARIABLE statement I added
GROUPING = part (0=groupA, 1=groupB);
This resulted in a warning message:
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE
COMPUTED. THE MODEL MAY NOT BE IDENTIFIED. CHECK YOUR MODEL.
PROBLEM INVOLVING THE FOLLOWING PARAMETER:
Parameter 28, Group GROUPB: [ DIFF ]
I am running out of ideas and I need help so solve this issue. I am grateful for any advice (eg giving me the correct syntax).
Best regards,
Gert