Title: A note on estimation of a mixture of multivariate Tobit models
Authors: Jan Bruha - Czech National Bank (Czech Republic) [presenting]
Abstract: The aim is to propose a technique to estimate the latent-class mixture of multivariate Tobit models using the approximate Bayesian computation. Using the computational experiments, we compare the proposed algorithm to alternatives, such as the Gibbs sampler. We apply the proposed algorithm to analyze latent classes of expenditure behaviour of Czech households.