Estimasi Parameter Model Regresi Linier dengan Pendekatan Bayes

Abstract

Two types of viewpoints in statistics are Frequentist and Bayesian Method. In Bayesian method sees a parameter as a random variable, so the value is not single. Frequentist method that are often used in linear regression are Ordinary Least Square (OLS) and Maximum Likelihood Estimation (MLE). But along with developments, several studies show the results of modeling that are better at using Bayesian method than the Frequentist method. The data used is Poverty data in 2017 from BPS East Kalimantan. The purpose of this study is to estimate the parameters of the regression model with the Bayesian method on data on the number of poor people and regional domestic products in East Kalimantan Province in 2017. To estimate the parameters of the Bayesian linear regression model it is used by the prior conjugate distribution. Then the markov chain is designed from the posterior distribution with Gibbs Sampler as many as 50.000 iterations and the estimated parameters that are the average of the Gibbs Sampler value are = 0.9149, = 5.462, and = 0.2827. From the Gibbs Sampler values ​​that have been obtained, a density function for each parameter is generated so that the Bayesian confidence interval (credible interval) for estimation is (0.85; 0.9836), (4.484; 6.439) and (0.2694 ; 0,296) for parameters .