2
$\begingroup$

I'm running a structural equation model using Stata's SEM builder and am having difficulty achieving convergence of even the most simple models with my dataset.

I have N=2544 independent observations. The simplified model in the image below fails to converge after thousands of iterations. I believe this suggests that the model is not identified (is that correct?). I have attempted including additional variables to improve model identification but but these have also failed to achieve convergence.

My primary question is, what are some general steps I can take to achieve model convergence when using SEMs? What aspects should I be thinking about? These aspects can be general (in terms of SEM theory) or specific to Stata.

enter image description here

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

This is an excellent question. For latent variables to be identified they typically require at least 3 indicators or other outcome variables associated with them. In your diagram you have the ACEs variable only having two outcome variables. It looks like you'd be better off just using the observed ACEs variable as a predictor of your latent Parenting variable.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.