Mixed Logit Model Estimation (Beginner)


#1

Hi All,

I have designed a discrete choice experiment to estimate WTP figures for a non-market good within a group where I think preference is heterogeneous. But I am currently struggling to estimate the model.

I have read Train’s “DCE Methods with Simulation”, but I am unable to implement the programming techniques he proposed for estimating the model parameters. I don’t have a programming background, but I can read and implement R command syntax. I have tried to read around but it’s not helping. Most of the resources I found were on general multinomial logit.

I just discovered this Forum today and I am new to Stan. I would be glad if someone can recommend a go-to Beginner Resource on Stan with step-by-step guide on how to implement the estimation procedure for this model. More interested in syntax-based guides.

Thanks for your support.


#2

Since you are comfortable with R I would suggest looking at the RStan vignette here which will show you how to fit a simple Stan model with data in R. Regarding the model you are interested in fitting, once you are familiar with using RStan, it sounds like you should look at the Multi-Logit Regression section in the Stan reference manual here. (The early chapters in the manual will help you understand the syntax of the Stan modeling language.)


#3

Thanks a lot.


#4

Also, this thread might be useful for you:


#5

Hey Aikins –

This might help!
https://khakieconomics.github.io/2018/03/13/Combined-conditional-and-multinomial-logit-in-stan.html

And this
https://khakieconomics.github.io/2018/03/05/Discrete-choice-in-stan-(sawtooth-con-talk).html

Let me know if you have any questions.

Cheers,
Jim


#6

Thank you very much for these wonderful resources. Will review them carefully and get back.


#7

I’m writing a chapter on this at the moment! Should be done in a few months. I’ll post it on this listserv and tag you when it’s up.