|
|
 |
| ESEM as measurement model in multigro... |
 
|
| Message/Author |
|
|
|
|
Dear Dr. Muthén,
I'm trying to compare structural weights across groups in a model which uses ESEM as measurement of the predictors (Big 5 factors) and a latent variable of Deviance as an outcome variable.
I wanted to keep the loadings free but constrain structural weights (B5 factors predicting deviance) to equality across groups, but I'm getting this fatal error:
*** FATAL ERROR
IMPROPER PARAMETER CONSTRAINT FOR EFA MEASUREMENT SPECIFICATION.
(Error Code: 1020)
Here's the code:
ANALYSIS:
Rotation = Target;
MODEL:
open by s20 ... (shortened for brevity) s44
s33~0 ... s03~0
s11~0 ... s16~0
s32~0 ... s07~0
s14~0 ... s29~0 (*1);
consc by s20~0 ... s44~0
s33 ... s03
s11~0 ... s16~0
s32~0 ... s07~0
s14~0 ... s29~0 (*1);
ext ...
agree ...
neuro ...
Dev by devP41 devP42 devP43 devP44;
Dev on open (b)
consc (c)
ext (d)
agree (e)
neuro (f);
Model CZE:
open ...
consc ...
ext ...
agree ...
neuro ...
Dev by ...
Dev on open (b)
consc (c)
ext (d)
agree (e)
neuro (f); |
|
|
|
|
If you remove this part of the model
Model CZE:
open ...
consc ...
ext ...
agree ...
neuro ...
you will be able to estimate the model, i.e., the model can be estimated with scalar or metric invariance where the loadings are held equal across the two groups.
For a more general constraint where the loadings in the esem group are not equal across groups but the regression parameters are you will have to switch to EwC (ESEM-within-CFA) approach described here
http://www.vanderbilt.edu/psychological_sciences/graduate/programs/quantitative-methods/quantitative-content/marsh_morin_parker_kaur_2014.pdf |
|
|
|
|
| Thank you for your tips. I will go over the paper as that's what I'm trying to do (leave loadings free and test equality of regression parameters across groups) |
|
|
| Back to top |
|
|
