A question about conditional logit model estimation in Biogeme

[Karine Xu]

Dear prof.Bierlaire,

I am estimating a conditional logit model which has not only the experimental attributes but also contains some social-demographic attributes of the decision-makers. My question is, does Biogeme require long-format or wide-format data in the conditional logit model estimation? Because I found the estimation results are different from the same model estimated in R mlogit, which makes me quite confused. 

If you have any suggestions and examples of the model containing social-demographic attributes please let me know.  

Kind regards,

Chenchen

[Bierlaire Michel]

Then, Biogeme is using the “wide” format: one row per observation. 

And, yes, you can include sociio-economic characteristics and context variables.  

On 12 Jan 2023, at 21:07, xuchen...@gmail.com wrote:

Dear prof. Bierlaire, 

Many thanks for the quick reply!

Sorry I didn't explain it clear. Examples of the long- and wide-data formats are shown below. The long-format data contains information for all alternatives. The examples are from https://www.gla.ac.uk/media/Media_835341_smxx.pdf 

<image.png>

<image.png>

I find many examples with Biogeme dealing with alternative-variant attributes, such as travel time and cost.  I want to consider the environmental attributes, such as the weather and the age of the decision-maker, which are not variant across the alternatives. I  would like to ask is there a way to take these attributes into the model in Biogeme please? 

Thank you for your advice!

Kind regards,

Chenchen Xu

On Thu, 12 Jan 2023 at 17:27, Bierlaire Michel <michel.b...@epfl.ch> wrote:

I am not sure what you mean by "long" or “wide”. See section 1 of the primer:

--
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/d94cc620-1739-47bc-96cd-c5062ecfecedn%40googlegroups.com.