The Cross-entropy Method for Estimation of Nonlinear Functions of Parameters in Bayesian VAR Models
Author | : Nuša Mikuljan Šljivić |
Publisher | : |
Total Pages | : 126 |
Release | : 2016 |
ISBN-10 | : OCLC:1005214413 |
ISBN-13 | : |
Rating | : 4/5 ( Downloads) |
Download or read book The Cross-entropy Method for Estimation of Nonlinear Functions of Parameters in Bayesian VAR Models written by Nuša Mikuljan Šljivić and published by . This book was released on 2016 with total page 126 pages. Available in PDF, EPUB and Kindle. Book excerpt: In this dissertation, an adaptive importance sampling method for estimation of the expected value of nonlinear function of parameters under the posterior density in Bayesian VAR model is developed. Most Bayesian inference problems can be expressed as the evaluation of the expectation of a function of interest, usually as a nonlinear function of the model parameters, under the posterior distribution. Nonlinear functions in Bayesian VAR setting are difficult to estimate and usually require numerical methods for their evaluation. In this dissertation, a weighted importance sampling estimator is used for the evaluation of the posterior expectation. The optimal importance sampling density, which minimizes the variance of the estimator, is in general difficult to evaluate and cannot be used in practice as the normalization constant of the density depends on the marginal likelihood. The proposed importance sampling approach uses cross-entropy method for determination of the importance sampling density. With cross-entropy approach the importance sampling density is chosen from a specified family of densities such that the cross-entropy distance or Kullback-Leibler divergence between the optimal importance sampling density and importance density is minimal. The performance of the proposed importance sampling algorithm with cross-entropy method is assessed in iterated multi-step forecasting of US macroeconomic time series.