Latent Variable after SEM

[Marco d'Errico]

Dear all

I have a question which relates to SEM and latent variable. I want to run a sem assuming that a number of dimensions (x1-x9) create a latent variable. in other words, I assume that a # of observed variables (which I have in my dataset) contribute in creating something that I don't have in my dataset.

Is there any way stata can do this? What I am sure stata does is to find the Betas between existing variables considering them as either latent or observed.

to give an example; i would like to write

sem(LATENT<- var1 var2 var3..var9), latent(LATENT)

and have my LATENT generated.

is it possible?


tks

[Clyde Schechter]

After running the -sem- there is a postestimation command -predict- which includes the option -xblatent-. So

predict, xblatent

will give you estimates of all your latent variables. -predict, xblatent(A B C)- will give estimates of just latent variables A, B, and C.

[Clyde Schechter]

Addendum to above: if your latent variables are not themselves endogenous, then you want -predict, latent- instead of -xblatent-.

[Marco d'Errico]

Clyde

thank you.