Adding syntax for default settings PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
 Dayuma Vargas posted on Friday, February 25, 2011 - 9:57 am

I am working on some invariance across time analysis. First I ran a simple CFA for one construct with 4 indicators at Time 1. When I ran the model, I received a warning that the model was not identified. I decided to add a syntax line specifying that the mean of the latent construct is fixed at zero. Once I ran the syntax again, the model worked fine and I got no warnings. I am not sure I understand why adding that line of syntax made a difference as I thought that it was the default in Mplus 6.1 for the mean to be fixed at zero to begin with.
 Linda K. Muthen posted on Friday, February 25, 2011 - 10:40 am
In a cross-sectional model the means of the factors are fixed at zero for model identification. Only in a multiple group or multiple time point model can factors means be identified in all but one group or all but one time point.
 Dayuma Vargas posted on Friday, February 25, 2011 - 11:24 am
Thank you for the quick response. If I run a model with the following syntax, where I ask for the means of the latent construct at each time point with the mean at time 1 fixed at 0, I receive an warning message stating that the model is not identified. However, if I remove the line [IDT1@0 IDT2 IDT3]; then there is no problem. But isn't the above line simply stating the same thing as the default settings in the program? So why should it make a difference whether I have that line in or not?

!Identity at Time 1
IDT1 by ID055* ID058 ID059@1 ID0511 ID0513 ID0514 ID0529;

!Identity at Time 2
IDT2 by ID061* ID062 ID063@1 ID064 ID065 ID066 ID0610;

!Identity at Time 3
IDT3 by ID075* ID078 ID079@1 ID0711 ID0713 ID0714 ID0729;

!Means & Intercepts
[ID055 ID061 ID075];
[ID058 ID062 ID078];
[ID059 ID063 ID079];
[ID0511 ID064 ID0711];
[ID0513 ID065 ID0713];
[ID0514 ID066 ID0714];
[ID0529 ID0610 ID0729];
[IDT1@0 IDT2 IDT3];

!Variances & Residual Variances
ID055 ID061 ID075;
ID058 ID062 ID078;
ID059 ID063 ID079;
ID0511 ID064 ID0711;
ID0513 ID065 ID0713;
ID0514 ID066 ID0714;
ID0529 ID0610 ID0729;

!Correlations & Covariances
IDT1 with IDT2 IDT3;
IDT2 with IDT3;
 Linda K. Muthen posted on Friday, February 25, 2011 - 1:12 pm
For the factor means to be identified, the intercepts must be held equal over time. See multiple indicator growth modeling in the Topic 4 course handout on the website. The inputs to test for measurement invariance across time are shown as part of this example.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message