Attachment | Size |
---|---|
Disentropy 2 steps SEM model.jpg [6] | 73.6 KB |
2_Step_Model_ Fit_Results.jpg [7] | 43.72 KB |
Hi,
I am making a 2-steps approach to estimate a SEM model in the field of economics as depicted in the attached figure. In order to obtain the estimated values of the latent variables observation-by-observation for N=142 countries in the 1st step I am using the predict function of lavaan.
Whilst fit results for the individual measurement models of the 1st step are pretty good, at the moment I have not got acceptable RMSEA and TLI in the 2nd step (RMSEA: 0.245, TLI: 0.897) as shown in the following chart:
RMSEA CFI TLI SRMR
1st step
Governance 0.077 0.996 0.991 0.014
Wealth 0.028 0.999 0.998 0.016
Conflict 0.071 0.996 0.988 0.014
Mutuality 0.067 0.991 0.983 0.029
2nd step
Disentropy 0.245 0.966 0.897 0.023
Optimality cut < 0.08 > 0.95 > 0.95 < 0.08
I have a set of measurable variables for each of the four latent variables and am trying to improve the overall model following a trial-and-error method.
-
Could you give me some guidance on how to proceed in order to respecify the 2nd step model?
-
Will the 2nd step model improve if the individual 1st step models improve?
-
Could I perform any sort of statistical tests / analyses on all the available measurable variables in order to set the 1st step individual models determining an acceptable fit of the 2nd step model?
Thanks
(Note: this thread follows a post I submitted to this fórum - General SEM Discussions - on 04/25/2015)
[img_assist|nid=4093|title=2-step SEM Model|desc=|link=none|align=left|width=90|height=100]