Advances and Applications in Statistics
Volume 43, Issue 1, Pages 37 - 52
(November 2014)
|
|
A BAYESIAN METHOD TO ESTIMATE THE PARAMETERS OF REGRESSION MODEL
Nasr Rashwan and Hanaa Salem
|
Abstract: This study presents the Bayesian method to estimate the parameters of regression model as an alternative method to the classical methods. Although, this method includes complex calculations but it is more accurate than other methods because it produces a direct probability statement about parameters and allows one to interpret a probability as a measure of degree of belief concerning the actual observed data. In Bayesian method, Gibbs sampling algorithm was used to select several iterations (samples) from posterior distribution of regression coefficients using noninformative prior and conjugate prior. The empirical results showed that the estimates of coefficients under the Bayesian regression model using conjugate prior are more accurate than the Bayesian regression model using noninformative prior. |
Keywords and phrases: Bayesian method, noninformative prior, conjugate prior, Gibbs sampling, posterior distribution. |
|
Number of Downloads: 345 | Number of Views: 1052 |
|