Discrete (categorical) input data

[Bluerrror]

Hello,

I am currently using the torchvinecopula library and have a question regarding the incorporation of categorical (discrete) input features into the model.

Is there a way to include discrete variables in the input data for the copula models? If so, could you please provide guidance or examples on how to properly encode or handle these categorical features within the framework?

Thank you for your help!

Best regards,
Shahab

[TY-Cheng]

Hi Shahab,

Thank you for reaching out! Currently the repo focuses on continuous variables. Modelling discrete data via copula need to address non-uniqueness and lack of identifiability (https://www.columbia.edu/~rf2283/Conference/1Fundamentals%20(1)Seagers.pdf).

I might not have the bandwidth to explore this fully at the moment—do you have experience or time to delve into this? I'd be happy to exchange ideas if you do.

Looking forward to your thoughts!

Best regards,
TY

[Bluerrror]

Hi TY,

Thank you for your reply. Actually, I dont have the time and the experience. However, in future I like to work on it. For others I can say I solved the issue of discrete variables as an input by using uniform distribution, however, I know it is not a perfect solution.

In this stage it is enough for me. However, there is many points which still could be sovled by using categorical input data.

Warm Regards,
Shahab