Hybrid model with Ordered Logit

[Yashin Abdullah Ali]

Dear Prof, 

I want to do a hybrid model with Ordered Logit. 

My choice of experiment design is generic and with unlabelled alternatives, which means I am estimating the generic coefficients. I would like to know, if it is possible to evaluate such a design for the complex model considering the explanatory variables such as income, age, gender, HHsize as a dummy variables? Whereas the measurement part of the model is an ordered outcome of the Likert scale 1-5. 

Specifying the structural model of the latent variables by including the socio-demographic characteristics and thereby including this latent variable in the utilities. 

Now, I am not sure to which of the alternatives should I incorporate the structural part of the latent variable? 

Should I incorporate to all three utilities? 

Note: the unlabelled alternatives are named: Option 1, Option 2, and Option 3. 

[Bierlaire Michel]

Everything is possible. But you first need to identify what your latent variables are actually capturing, and what are the indicators that allow its measurement. 

Then, you can include them in your model as if ithey were observed variables, thanks to the structural equations. 

See https://transp-or.epfl.ch/documents/technicalReports/Bier18b.pdf

--
You received this message because you are subscribed to the Google Groups "Biogeme" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biogeme+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/biogeme/7a67961b-ac5e-471f-affd-bef9a2fb9710n%40googlegroups.com.

[Yashin Abdullah Ali]

Dear Sir,

Thank you for your reply.

Is it necessary to develop more than one HCM model and compare both of them through AIC fit?

Or just one HCM model answers the question of the latent variable’s?

[Bierlaire Michel]

HCM is like any other model. You may want to test different specifications, to identify the one that represents best your data.

To view this discussion on the web visit https://groups.google.com/d/msgid/biogeme/6ef94b9f-ff68-45ce-86bf-0f22e84ac8b3n%40googlegroups.com.