Advances and Applications in Statistics
Volume 40, Issue 2, Pages 157 - 167
(June 2014)
|
|
ENSEMBLE TWO RETURN LEVELS OF GENERALIZED PARETO AND MODIFIED CHAMPERNOWNE DISTRIBUTIONS USING LINEAR REGRESSION
A. H. Wigena, A. Djuraidah and I. W. Mangku
|
Abstract: This paper is concerned with weighted ensemble to combine two return levels as extreme values resulted from generalized Pareto and modified Champernowne distributions. Both the distributions are used to predict extreme values. However, the first distribution results in over estimate while the second results in under estimate. We propose a technique to find weights based on linear regression without intercept to combine both the estimates. Our results show that the ensemble technique using simple linear regression gives exactly the same weights as the trial and error technique using iterative process. |
Keywords and phrases: ensemble, extreme, generalized Pareto, modified Champernowne, return value, regression. |
|
Number of Downloads: 347 | Number of Views: 1125 |
|