Posterior analysis of stochastic frontier models with truncated normal errors (Q1861638)

The regression model \(y_i=x_i^T\beta+\nu_i-u_i\), \(i=1,\dots,n\), is considered, where \(y_i\) is the response, \(x_i\) is the vector of regressors, \(\beta\) is the vector of unknown parameters, \(\nu_i\sim N(0,\sigma^2)\), and \(u_i\) is distributed according to the normal distribution \(N(\mu,\omega^2)\) truncated below at zero. The problem is to estimate the parameters \(\beta\), \(\sigma\) and \(\omega\). The Bayesian approach to this problem is considered. Posterior inference using the Gibbs sampler is described. Results of simulations and applications to airline production function data are presented.