Restricting residual variances to be ... PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 Vika Savalei posted on Tuesday, April 18, 2006 - 11:14 am
I am trying to run a 1-factor model with one of the residual variances constrained to remain positive.

I've consulted example 5.20 in the new manual, and based on it set up the following:

VARIABLE: NAMES ARE y1-y5;
MODEL: f1 BY y1-y5 (V1-V5);
MODEL CONSTRAINT:
V5 > 0;

This runs, but the output is the same as if the last two lines were never there. What am I doing wrong?

Thanks much!
 Linda K. Muthen posted on Tuesday, April 18, 2006 - 11:57 am
v5 refers to a factor loading not a residual variance. To refer to a residual variance, say:

y1-y5 (v1-v5);

and remove the labels from the factor loadings. If you continue to have problems, send the input, data, output, and your license number to support@statmodel.com.
 Vika Savalei posted on Tuesday, April 18, 2006 - 10:03 pm
Hi Linda,

Thanks for your message. I'm not sure I understand how to modify the syntax. I already say: f1 on y1-y5 (v1-v5), except I use capital V's. Should I repeat this statement again with lower case v's as you suggest? Also, what does it mean, "remove labels from factor loadings"?
 Linda K. Muthen posted on Wednesday, April 19, 2006 - 6:35 am
You refer to residual variances by the name of the variable alone. ON is for regression coefficients. BY is for factor loadings. Try

VARIABLE: NAMES ARE y1-y5;
MODEL: f1 BY y1-y5;
y1-y5 (V1-V5);
MODEL CONSTRAINT:
V5 > 0;
 John D Peipert posted on Monday, February 24, 2020 - 11:51 am
Hi -

I am trying to run a bifactor CFA with categorical indicators using theta parameterization. I want to label the residual variances so that I can use them to calculate omega hierarchical in the Model Constraint area. However, when I add the term for the residual variances to the model, I get an error that the residual variance matrix may not be positive definite and the model may not be identified. Can you help? Thanks!

MODEL:
GEN BY y1* y2-y8 (L1-L8);
F1 BY y1* y2-y3 y6 y8 (L9-L13);
F2 BY y4* y5 y7 (L14-L16);

GEN WITH F1-F2@0;
F1 WITH F2@0;

GEN@1;
F1-F2@1;

y1-y8 (R1-R8);
 Bengt O. Muthen posted on Monday, February 24, 2020 - 5:12 pm
The residual variances are not identified. Omega is not defined for categorical outcomes. See the FAQ on our website:

Reliability - binary and ordinal items
 John D Peipert posted on Wednesday, February 26, 2020 - 10:04 am
Thank you very much for your quick response!
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: