|
|
 |
| Covariance pb in CFA of categorical w... |
 
|
| Message/Author |
|
|
|
|
Dear Linda or Bengt,
I would like to run a CFA on categorical data using MLR link probit.
In the model, I specify the residual errors of 2 observed variables to covary. I run this model already with WLSMV (delta parametrization) without problem. But the MLR analysis brings this error message:
"*** ERROR in MODEL command
Covariances for categorical, censored, count or nominal variables with
other observed variables are not defined. Problem with the statement:
V4 WITH V5"
Here is my model syntax:
VARIABLE: NAMES ARE V1 V2 V3
V4 V5 V6 V7;
CATEGORICAL IS ALL;
MISSING IS ALL (999 998 997);
MODEL: Aff BY V1 V2;
Cog BY V3 V4
V5 V6 V7;
V4 WITH V5;
OUTPUT: standardized;
modindices (ALL);
ANALYSIS: ESTIMATOR=MLR;
LINK=PROBIT;
PLOT: TYPE IS PLOT2;
Would you mind helping me understand what is wrong?
Many thanks! |
|
|
|
|
| With maximum likelihood estimation and categorical outcomes residual covariances are not part of the model. Each residual covariance constitutes one dimension of integration. You need to specify these using the BY option as shown in Example 7.16. The ease of estimating residual covariances is one of the strengths of weighted least squares estimation with categorical outcomes. |
|
|
| Back to top |
|
|
