Numerical analysis for switching regression models in a classical and a Bayesian framework

This paper deals with the numerical problems which are attached to the statistical analysis of switching regressions model either from a classical or a Bayesian point of view. For the first case are detailed an algorithm for the optimization of a non-smooth likelihood function and a procedure for choosing starting values. For the second case the interest is focused on Monte Carlo methods for integrating the posterior density. The paper is illustrated by economic examples devoted to the labour market.