ApolloChoiceModelling forum

Dear Prof. Hess and Dr. Palma

As a part of my new study, I need to estimate a mode choice model for various trip purposes based on RUM, regret minimization and cumulative prospect theory. Based on my understanding the latter one allow for the inclusion of psychological factor related loss aversion, probability weighting and reference depdendence.

I would request you if it is possible to suggest the steps/codes to apply this model in apollo by utilizing the mode choice example by adding some dummy values. Also are there any specific questions that are needed to be included in the survey for parameter weights in CPT.

Hi

can you please be a bit more specific of what you want to specify, as this is very vague. Do you mean you want to allow for gains-losses asymmetry? Are you still wanting to do this in a RUM context or not? If you want to use RUM, you can specify separate parameters for gains and losses in the utility function, for example

Stephane

Thankyou Prof. Hess for prompt response.

To simply put, I want to develop a discrete choice model by utilizing cumulative prospect theory framework. As per my understanding to develop such a model we need additional data points like

1. Probablity Weighting Function
2. Value Function
3. Reference point
4. Loss aversion

I wanted to know if it is possible in Apollo to develop a choice model based on CPT?
My request is if you can provide code for developing such a model by assuming dummy values for these four data points in the mode choice example in Apollo.



For my reference I have studied the following papers.

1. https://doi.org/10.1016/j.tra.2013.01.011
2. https://doi.org/10.1016/j.tra.2022.103564

Hi

sorry, we don't have time to produce specific code for this. What you need to determine is whether you can approximate this inside a RUM model, in which case you can use existing functions. An example of such a specification for gains-losses is given by https://doi.org/10.1016/j.tre.2007.06.002

If it's a case of a completely different model, then you could still use Apollo, but you would need to write the probabilities yourself inside apollo_probabilities

Sorry to not be able to be of more help, but we need to prioritise key developments

Stephane