Performance Bayesian quantile regression for modeling poverty line with Violation of Heteroscedasticity Assumption
Lilis Harianti Hasibuan (a), Ferra Yanuar (b*), Dodi Devianto(b), Maiyastri(b)

a) Faculty of Science and Technology Imam Bonjol State Islamic University Padang
Jl.Sungai Bangek, Koto Tangah, Padang, Indonesia, 25153.
lilisharianti[at]uinib.ac.id
(b) Faculty of Mathematics and Natural Sciences, Andalas University
Kampus Limau Manis, Padang, Indonesia, 25163.
ferrayanuar[at]sci.unand.ac.id

Abstract

Bayesian quantile regression is a parameter estimation method that combines the concept of quantile analysis into the Bayesian approach. This study aims to implement the Bayesian quantile regression method in modeling the poverty line in West Sumatra. The poverty rate data violates one of the classical assumptions, namely that the data contains heteroscedasticity. This study considers the Bayesian quantile regression approach using the Gibbs sampling algorithm of the Bayesian method for fitting the quantile regression model. This research explores the performance of asymmetric Laplace distribution for working likelihood in posterior estimation process. The results showed that the best model for the poverty rate in West Sumatra was in the 0.5 quantile by producing an MSE value of 0.217

Keywords: Bayesian quantile regression- poverty line- Gibbs sampling- MSE

Topic: Statistics