Work

Bayesian Heteroscedasticity Model With Variable Selection Method

Year: 2015
Type: dissertation
Abstract: Homoscedasticity is an important assumption in the application of linear regression analysis.However, heteroscedasticity, the absence of homoscedasticity, is often the case with real data, such as time-series data, panel data or cross-section data.In order to deal with this problem, different methods are suggested by the classical statistics based on the knowledge of the variances' structure or at least estimation of the structure.In the paper, a hierarchical Bayesian model with stochastic search variable selection method (SSVS) is built in order to estimate parameters and the structure of the heteroscedasticity at the same time.Markov Chain Monte Carlo(MCMC) simulations like the Gibbs sampling and the Metropolis-Hasting sampling are applied to get the posterior distributions for the parameters and the SSVS method is evaluated with regards to identify the structure of heterosedasticity.
Author: Wei Wei Feng
Institution: Örebro University
Language: English
Open Access status: green
OpenAlex ID: W568590379
Cites: 11
Cited by:
Related to: 10