Question: SEM with latent predictors and observed DV

[Miranda Baumann]

Hello all,

I apologize for the stupid question, but I am having a difficult time with (a) running a latent path analysis, and/or (b) interpreting what I'm running.

I am attempting to conduct a path analysis of two latent variables (Religiosity [2 var's] and School attitudes [4 var's]) predicting a continuous observed DV (alcohol use). The hypothesis is Religiosity -> School -> alcuse. I have tried running this with the SEM builder and by typing out commands, but I am struggling.

In a perfect world, I would want to be able to assess both the direct and indirect effects of religiosity on alcohol use by not constraining the model through School only, but I can't figure out how to do that. Also, I think my commands have the effect of making alcohol use one of the variables making up School. I suppose this makes practical sense, but I'm not sure what I'm doing.


Would someone mind taking a look at what I'm doing and tell me what I'm doing wrong? I can't find the direct and indirect effects anywhere in the output or using estat teffects command, which leads me to believe I'm doing something wrong.


Thanks!!

~ Mir


I have tried the following command syntax:

sem (Religiosity -> relimp reldec) (School -> feltschl mngflschl learnimp intschl) (Religiosity -> School) alc30, stand


I have also tried building the following models via SEM builder:

[Clyde Schechter]

The first diagrammed model looks appropriate for your purpose to me. To get a sense of the direct vs indirect contributions of Religiosity to alc30, following that -sem- you can run -estat teffects-.

[Miranda Baumann]

When I ran the teffects command, my outcome was listed under the measurement model, not the structural one. That's what was confusing me.

[Clyde Schechter]

I don't think that matters.

[Miranda Baumann]

Thank you very much!