ApolloChoiceModelling forum

Hi David and Stephane!

I have a question about price variable.

In my stated preference database, I have four levels (150, 250, 350, 450) for the price attribute.

Is that a correct practice, if I divide by 100 these levels in my database and I use these levels (1.15, 2.5, 3.5, 4.5) for estimate models?

I've experienced it before, price attribute levels with narrower ranges work better, especially in case of random parameter models, and models is WTP-space.

And after, when I calculate WTP estimates (beta(attribute)/beta(price)), I multiply by 100 the received values.

Is it a correct practice?

Could you recommend literature on the subject?

Thank you very much for your answer!

Best regards,
David

David

changing the units of the price attribute in this way will just change the units of the coefficient, so your model will remain the same. Rescaling like this can be beneficial sometimes for computational reasons - have a look at the discussions in the manual about using scaling, which means you don't need to do the transformations yourself

Stephane

Thank you for your help Stephane!