|
|
 |
| |
|
|
|
Comparison of the Estimation Efficiency of Regression Parameters Using the Bayesian Method and the Quantile Function |
|
|
|
PP: 11-20 |
|
|
|
doi:10.18576/jsapl/060102
|
|
|
|
Author(s) |
|
|
|
Ismail Hassan Abdullatif Al-Sabri1,2,
|
|
|
|
Abstract |
|
|
| There are several classical as well as modern methods to predict model parameters. The modern methods include the
Bayesian method and the quantile function method, which are both used in this study to estimate Multiple Linear Regression (MLR)
parameters. Unlike the classical methods, the Bayesian method is based on the assumption that the model parameters are variable, rather
than fixed, estimating the model parameters by the integration of prior information (prior distribution) with the sample information
(posterior distribution) of the phenomenon under study. By contrast, the quantile function method aims to construct the error probability
function using least squares and the inverse distribution function. The study investigated the efficiency of each of them, and found that
the quantile function is more efficient than the Bayesian method, as the values of Theil’s CoefficientU and least squares of both methods
came to be U = 0.052 and åe2i
= 0.011, compared to U = 0.556 and åe2i
= 1.162, respectively. |
|
|
|
|
|
|
|
