Choosing the right model for my data - Choice model

[Jean Dewals]

Dear collegues,

I'm currently working on the result of a consumers surveys. In this survey, we asked to our individuals to rank their preferred labelling scheme among a pool of ten alternatives. These labels concern seafood products.

My objective is them to analyse the consumption profiles behind the expressed preferences regarding these 10 alternatives according to our consumers characteristics.

To do so, i started by using a Mixed logit modelthrough the following command in STATA

cmset ID alternatives

(ID) to identify my consumers and (alternatives) my 10 schemes

Then i use:

cmmixlogit rang, casevars(i.recode_age i.genre deplittoral prix_crit enfant univ tradi prefgeo) basealternative (10)

(Rang) to dentifies the alternative ranked 1st in preferences
After, i have only case-specific variables (age, gender, place of residence, presence of children in the household....).
I don't have any random parameter in my code. I also don't have any alternative-specific variable, my labels being defined solely by their name



My question is the following : Knowing that Mixed logit are usefull to estimate random parameters, is this the best model to use for my data?

At first, i chose to use this model for two principal reasons:

- The IIA assumption
- The fact that i want to compare the preferences expressed between them.

But now, after several discussion, i'm not sure that this king of model fits the best with my data and my purpose.

I don't know if you have enough information to answer my question, but I remain available if you need me more.

Best regards,

Jean-François

[William Liu]

I'm not particularly familiar with consumer choice models, but I thought I might chip in to help.

It seems like you're looking for a rank-ordered consumer choice model to take advantage of the fact that your data fully captures preference rankings rather than just the top choice; using a model without rank-ordering discards information.
You might be interested in the commands -cmrologit- and -rologit-. (The latter is the older counterpart of the former.)

These estimate the Allison and Christakis (1994) model. I don't think that model allows the coefficients to be heterogeneous amongst consumers, which I presume you might be looking for since that's a core part of -cmmixlogit-. There probably isn't a ready-made Stata command to estimate rank-ordered consumer choice models with coefficient heterogeneity, though you could try manually programming such a procedure based on said model. One such model I found is: https://core.ac.uk/download/pdf/18515565.pdf.

[Jean Dewals]

Thanks for your answer