ApolloChoiceModelling forum

I am looking to estimate a coordinated daily activity pattern choice model. The specific likelihood function might be a mixed logit or latent class, but I am curious if there are methods in Apollo to set up the coordination explicitly.

Specifically, consider two users n and m, choosing alternative j among choice set J. The utility function for an alternative j n is a function of their own utility, the other user's utility for that option, and any variables that might describe the interaction.

U_jn = U_jn + U_jm + X_jnm\lambda

An example application of this model is Gupta, et al. (2006) https://journals.sagepub.com/doi/abs/10.3141/2493-06. One known way to estimate this model is to expand the choice set from J to J^N, where N is the number of individuals that get coordinated. But this explodes real quick when there are multiple members of a household, or the number of alternatives is high.

Is it possible to configure the utility functions or likelihoods in Apollo to estimate this model without extensive data preparation? That is, could we specify a variable to identify coordination and then specify joint variables?

Hi Gregory,

Currently, Apollo does not include any convenience routine to implement what you need. The closest I could recommend is using a loop to define the alternatives' utility in a more efficient way, in a similar fashion to example 30 in the website (http://www.apollochoicemodelling.com/fi ... ample_30.r).

Best
David