MIMIC model

[Marjan Naumoff]

Hi everyone

I am reading a bunch of literature on MIMIC models for the first, but I am a bit confused. Do I have a problem if some of the 'causes' is endogenous (simultaneous) to the latent variable? Or, the causality from the 'cause' to the latent variable is established by construction?

Thank you for your consideration.
Marjan

[Maarten Buis]

Can you give us the exact model you have in mind?

[Marcos Almeida]

No doubt, as Maarten underlined, you are bound to get better advice provided you present details of your model.

That said, you may wish to take a closer look at the difference between reflective and formative indicators.

[Marjan Naumoff]

Hi. Thank you.

This is what I want to do:

sem (SubjSES -> q98a q98c) (SubjSES <- $individual $household unemployed) [pw=weight], vce(robust)

q98a and q98c refer to opinions on mental health (cognitive skills, memory etc.). So, my latent variable is 'mental health'. This is explained by a set of individual and household characteristics, plus whether the person is employed or not.

The problem is the simultaneity between mental health and unemployment: a person may become unemployed and then his/her mental health sees deterioration; or the person may be already mentally ill which causes unemployment.

In standard regressions, we attempt to this by employing exogenous instrument, affecting unemployment but not mental health directly.

My dilemma in MIMIC is if the causality is assumed by construction or not? If not, what could be done?

Thank you again.

[Marjan Naumoff]

Can you give us the exact model you have in mind?

Hi Maarten. I posted above the more detailed question. I would appreciate if you could help. Thank you.