| Message/Author |
|
|
|
|
Hello,
I am examining the relation between a CFA measurement model for depression and several binary and continuous predictors using WLSMV estimation. One of my factors consists of only 2 indicators. When I fixed the factor loadings of the 2 indicators to be equal, the model did not converge. When I fixed both the factor loadings to be equal and the variance of the factor to 1, the model converged. Question: Is it correct to apply both these constraints? Or should I just constrain the factor variance to 1 and freely estimate the 2 factor loadings?
Many thanks in advance.
Model:
Dep by bcesd03 bcesd06 bcesd09 bcesd10 bcesd14 bcesd17 bcesd18;
Pos by bcesd04 bcesd08 bcesd12 bcesd16;
Som by bcesd01 bcesd02 bcesd05 bcesd07 bcesd11 bcesd13 bcesd20;
Int by bcesd15* bcesd19 (eq1) ;
Int @1;
Dep on Bcode sex aedu bmage xnochrot bmalcnei bmmedad bmbenanx bpartner
bal_dia B_CVD sex aedu bmage btotact smoke;
Pos on Bcode sex aedu bmage xnochrot bmalcnei bmmedad bmbenanx bpartner
bal_dia B_CVD sex aedu bmage btotact smoke;
Som on Bcode sex aedu bmage xnochrot bmalcnei bmmedad bmbenanx bpartner
bal_dia B_CVD sex aedu bmage btotact smoke;
Int on Bcode sex aedu bmage xnochrot bmalcnei bmmedad bmbenanx bpartner
bal_dia B_CVD sex aedu bmage btotact smoke; |
|
|
|
|
| The model is not identified by holding loadings equal - you must fix the factor variance at one. |
|
|
|
|
Thank you. Are you saying that I should:
(a) fix the factor variance to 1, as well as hold the factor loadings equal? or
(b) set the variance to 1 and freely estimate the factor loadings?
In both cases the model is identified - but how does the interpretation of the results differ in one case vs. the other? |
|
|
|
|
| Either way identifies the model as you say. Choosing to hold loadings equal or not depends on theory and model fit. You should have a substantive reason for holding them equal - otherwise don't. |
|
|
| Back to top |
