DENOISING LINEAR COMBINATION OF ORDER STATISTICS

Abstract

This study is concerned with the problem of Denoising Linear Combination of Order Statistics. Linear combination of order statistics termed L-Estimator has been considered in the literature, under regular condition this estimator is known to be asymptotically normal. Gaussian, Epanechnikov, Triweight, Triangular and cosine kernel have been used as the smoothers and various bandwidth selections were considered for the kernels. The result of the various denoised linear combinations of order statistics were compared with undenoised linear combination of order statistics using Mean Square Error (MSE). The results show that various denoised linear combination of order statistics are better than undenoise linear combination. Moreover the results have shown that denoising with Epanechnikov kernel under regulrar condition is better than denoising with every other kernel.