A0858
Title: Modelling preferences via Wallenius process
Authors: Veronica Ballerini - University of Florence (Italy)
Rosario Barone - University of Rome Tor Vergata (Italy) [presenting]
Brunero Liseo - Sapienza Universita di Roma (Italy)
Abstract: The volume of sales in a local real estate market is well known to be subject to fluctuations that depend on quotations, that depend on socioeconomic macro variables in turn. However, there exist micro and less volatile determinants of such volume of sales, that are based on the aggregate individuals' preferences. To disregard such determinants would imply returning biased predictions on the relative volume of sales for different municipality areas, conditioned on the observed quotations. In fact, buyers' preferences with respect to different zones of the local real estate market can be estimated assuming that the periodical sales volumes follow a Wallenius noncentral hypergeometric distribution (WNC), and that a sequence of WNC generates a newly defined stochastic process, i.e., a Wallenius process (WP). WNC describes a biased urn problem in which the probability to sample a certain number of colored balls depends not only on the number of balls of that color in the urn, but also on the weight associated with each color. Given the intractability of the likelihood function, the inference is performed via Approximate Bayesian computation (ABC) methods.
