ApolloChoiceModelling forum

Hello, everyone

I have a very simple question on structural equation part of the hybrid model. While specifying the structural part of latent variable as random coefficients is it fine to just use the normal draws.

For example
I have interNormalDraws = c("eta")

## Now in random coefficient
I have randcoeff[["LV"]] = eta. There are separate indicator equations but the structural equation just has variance.
Basically, I just want to see the effect of latent variable (LV) constructed from different items (indicators) in the choice model, does the way of defining model as mentioned have some complications in model estimation or explanation?

Hi,

Your model definition is fine. Your measurement equations will anchor and provide an interpretation for your latent variable, so you can still estimate your model and interpret its results.

Without any explanatory variables in the structural equation you won't be able to explain why the LV is high or low, or what characteristics give an individual a higher or lower LV. But if you are not interested in explaining that, then there is no problem with your formulation.

Cheers
David