Bayesian model averaging using power-expected-posterior priors

Dimitris Fouskakis and Ioannis Ntzoufras

This paper focuses on the Bayesian model average (BMA) using the power-expected-posterior prior in objective Bayesian variable selection under normal linear models. We derive a BMA point estimate of a predicted value, and present computation and evaluation strategies of the prediction accuracy. We compare the performance of our method with that of similar approaches in a simulated and a real data example from economics.

MoreLess

Year of publication:	2020
Authors:	Fouskakis, Dimitris ; Ntzoufras, Ioannis
Published in:	Econometrics : open access journal. - Basel : MDPI, ISSN 2225-1146, ZDB-ID 2717594-7. - Vol. 8.2020, 2/17, p. 1-15
Subject:	Bayesian model averaging \| Bayesian variable selection \| expected-posterior priors \| imaginary training samples \| power-expected-posterior priors \| Bayes-Statistik \| Bayesian inference \| Theorie \| Theory \| Prognoseverfahren \| Forecasting model \| Statistische Methodenlehre \| Statistical theory \| Schätzung \| Estimation