Advances and Applications in Statistics
Volume 57, Issue 2, Pages 155 - 189
(August 2019) http://dx.doi.org/10.17654/AS057020155 |
|
A NEW DIFFERENCE METHOD FOR VARIANCE ESTIMATION IN NON-PARAMETRIC REGRESSION MODELS
Magda M. M. Haggag
|
Abstract: In this paper, a new method for estimating the residual variance in nonparametric regression is proposed. This method is a new version of difference-based estimators of the residual variance, and called the difference-for-difference (DFD) estimation method. The DFD variance estimator is an unbiased estimator of the residual variance for difference order greater than one. Simulation studies are conducted to compute and evaluate the proposed variance estimator with other exiting estimators. Best results are obtained in the form of less relative mean squared error (RMSE) of the DFD variance estimator compared to other difference estimators. The results showed that the proposed estimator performs better than competing estimators in a large number of cases. |
Keywords and phrases: difference-based estimation, difference order, difference sequence, efficiency, nonparametric regression, residual variance estimation.
|
|
Number of Downloads: 328 | Number of Views: 986 |
|