Jean Frisou posted on Tuesday, November 28, 2006 - 5:38 am
In a structural equation model, how can I constrain the variances of dependant latent factors to be equal ? Indeed, the statements “ f1 (1); f2 (1);” constrain the variances of f1 and f2 to be equal, only if f1 and f2 are independant variables and constrain the residual variances of f1 and f2 to be equal if f1 and f2 are dependant.
You cannot constrain variances for dependent variables in the MODEL command because residual variances are estimated as part of the model for dependent variables not variances. Why do you want to do this?
Jean Frisou posted on Tuesday, November 28, 2006 - 10:06 am
Thank you Linda for your answer. I want to test equality of the standardized coefficient of the same relation, in two groups. If the variances of factors and raw estimate of the coefficient are respectively equal beteween group, then the standardized coefficient estimates are equal. Model G1 f2 on f1 (1); var f1 (2); var f2 (3); Model G2 f2 on f1 (1); var f1 (2); var f2 (3);
I think a better way to do this is to use MODEL CONSTRAINT to create standardized coefficients which you can then hold equal. See Example 5.20.
Jean Frisou posted on Wednesday, November 29, 2006 - 3:22 am
Thank you very much Linda for your suggestion. But I have a question. Does the label “f (vf);” refer to residual variance or variance factor ? In a CFA (example 5.20) f is an independant latent factor and in my model f is a dependant variable.
If your variable is depednent, then the label refers to the residual variance. You would need to create a new variable in model constraint that is the variance and use that for the standardization.
Jean Frisou posted on Thursday, November 30, 2006 - 1:13 am
Thank you for your help. I have created a variable with constraint model. My data are categorical and I use wlsmv. For the test, I use difftest method. But the output file indicates “DIFFTEST is not available in conjuction with nonlinear constraints through the use of MODEL CONSTRAINT. Request for DIFFTEST is ignored.”. In order to realize manually the test with the webnote of Bengt Muthén and Tihomir Asparouhov, could you say me how the structure of the difftest file is. There is only one column and it is not easy to find the matrix M.