Discrete Choice Model for Solar Panel Mode Choice

[Karloz Ferdinand Yco]

Dear Michel,

I am a masters student in sustainability and I want to do a discrete choice model on system choice for solar photovoltaic panels. I tried to use nested logit model having 2 modes: Mono (Monocrystalline), and Port (Portable). Mono has another nest containing (Mono 1) an (Mono 2). (Refer to the image below)

I ran the code successfully however my results always indicate that "Warning: Identification Issue" or "Algorithm failed to converge".

I have been trying for quite some time but couldn't generate proper results. Is there any way this can be fixed?



Please check on the code and my set of data that is attached in this email for reference.

Thank you.

Regards,
Karloz Yco

[Michel Bierlaire]

Try to estimate a logit first, and come up with a specification of the utility function that you are comfortable with. Once it is done, you can start testing some nesting structure.
It will allow you to see if the problem comes from the data, the specification, or the nested logit.

> On 5 Jun 2024, at 17:23, Karloz Ferdinand Yco <karl...@gmail.com> wrote:
>
> Dear Michel,
>
> I am a masters student in sustainability and I want to do a discrete choice model on system choice for solar photovoltaic panels. I tried to use nested logit model having 2 modes: Mono (Monocrystalline), and Port (Portable). Mono has another nest containing (Mono 1) an (Mono 2). (Refer to the image below)
>
> I ran the code successfully however my results always indicate that "Warning: Identification Issue" or "Algorithm failed to converge".
>
> I have been trying for quite some time but couldn't generate proper results. Is there any way this can be fixed?
>

> <445377378_2262243757455300_3768657612941293135_n.jpg>
>
> Please check on the code and my set of data that is attached in this email for reference.
>
> Thank you.
>
> Regards,
> Karloz Yco
>
>

> --
> 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/2562c3b9-bd08-48ec-843b-51ad290bf099n%40googlegroups.com.
> <data_updated_v10.2.csv><445377378_2262243757455300_3768657612941293135_n.jpg><20240604.py>

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

[Karloz Ferdinand Yco]

Hi Michel,

Thank you for your response. Following your recommendation, I proceeded with using a logit model for my research. Unfortunately, I am still encountering the same error. I am starting to suspect that the issue might be related to the dataset. Do you think a sample size of 106 is adequate?

Best regards,
Karloz Yco

[Michel Bierlaire]

> On 6 Jun 2024, at 17:31, Karloz Ferdinand Yco <karl...@gmail.com> wrote:
>
> Hi Michel,
> Thank you for your response. Following your recommendation, I proceeded with using a logit model for my research. Unfortunately, I am still encountering the same error. I am starting to suspect that the issue might be related to the dataset.

So, you have narrowed down the problem.

> Do you think a sample size of 106 is adequate?

It depends on the model, and on the variability in the data.
Start with simple specifications, with only a few variables.

> To view this discussion on the web visit https://groups.google.com/d/msgid/biogeme/41ea1420-6d3e-4abd-9978-9b07b4778e55n%40googlegroups.com.