Discrete Choice with Probabilistic Inversion : Application to energy policy choice and wiring failure

Discrete choice models normally represent the utility values of a finite set of mutually exclusive and exhaustive choice alternatives in terms of an expected value and an error. Strong assumptions are introduced on the error term. A different approach is sketched here. We assume that each member of a population of decision makers has a utility function, and that these functions can be scaled to have the same 0 and 1 values. Starting with a uniform distribution over the set of such utility functions, we apply probabilistic inversion techniques to acquire a distribution over the set of utility functions which optimally complies with discrete choice data. More generally, we may have some model explaining the discrete choice, where the parameters of the model are distributed over the population of actors. Probabilistic inversion finds an optimal distribution over the model parameters, given a starting distribution.