MXL Model with Sociodemographic data

[Georgia Charalampidou]

Dear Prof Bierlaire and forum members,

I am having problems with my MXL model when I try to introduce interactions with other variables. My dataset is a panel dataset, with each respondent occupying six rows. 

I firstly wanted to explore the effect of different variables on the model fit. To do this, I estimated an MNL, without accounting for the panel effect. The following code was used to introduce the segmentation (for the effect of income), and the estimation was executed.

inc_segmentation = DiscreteSegmentationTuple(
    variable=Variable('GENDER'),
    mapping={0: 'MAN', 1: 'WOMAN'}
)

When I switched to the MXL model and started checking for interactions with sociodemographic variables, I get the following error:

I tried introducing segmentation in different ways, and I'm sure the model specification is correct because the estimation process runs smoothly when interactions are not checked for. 

Do you have any ideas on how to solve this problem?

Thank you in advance for your help!

Kind regards,

Georgia Charalampidou

[Michel Bierlaire]

Georgia,

This was a bug in Biogeme. It is now corrected.
Update your version of biogeme using

uv add git+https://github.com/michelbierlaire/biogeme.git

or

pip install git+https://github.com/michelbierlaire/biogeme.git

Let me know if you still have problems.

Michel

> On 27 Mar 2026, at 13:16, Georgia Charalampidou <georgiachar...@gmail.com> wrote:
>
> Dear Prof Bierlaire and forum members,
>
> I am having problems with my MXL model when I try to introduce interactions with other variables. My dataset is a panel dataset, with each respondent occupying six rows.
>
> I firstly wanted to explore the effect of different variables on the model fit. To do this, I estimated an MNL, without accounting for the panel effect. The following code was used to introduce the segmentation (for the effect of income), and the estimation was executed.
>
> inc_segmentation = DiscreteSegmentationTuple(
> variable=Variable('GENDER'),
> mapping={0: 'MAN', 1: 'WOMAN'}
> )
>
> When I switched to the MXL model and started checking for interactions with sociodemographic variables, I get the following error:
>

> <Screenshot 2026-03-26 133050.png>
>
> I tried introducing segmentation in different ways, and I'm sure the model specification is correct because the estimation process runs smoothly when interactions are not checked for.
>
> Do you have any ideas on how to solve this problem?
>
> Thank you in advance for your help!
>
> Kind regards,
> Georgia Charalampidou
>
>
>

> --
> 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 visit https://groups.google.com/d/msgid/biogeme/0e638a00-797c-4296-acdc-1c2109cd8d33n%40googlegroups.com.
> <Screenshot 2026-03-26 133050.png>

Michel Bierlaire
Transport and Mobility Laboratory
School of Architecture, Civil and Environmental Engineering
EPFL - Ecole Polytechnique Fédérale de Lausanne
http://transp-or.epfl.ch
http://people.epfl.ch/michel.bierlaire