A Study on the Linear Estimation of Scale and Location Parameters by Order Statistics

Abstract

順序統計量의 一次結合으로 이루어진 不偏推定量이 最良線型不偏推定量이 되는 경우는 많이 있다. 그러나 順序統計量에 의한 어떤 推定量은 最良不偏推定量이 아닐수가 있다. 비록 順序統計量의 一次結合으로 된 모든 推定量들이 最良不偏推定量보다 有效性이 떨어지지만 그러한 推定量을 사용할 때 많은 잇점이 있는 경우가 있다. 本 論文은 順序統計量에 의한 보다 나은 線型推定量들이 一般化한 最小自乘推定法으로 얻어질 수 있음을 제시하였다.

In many cases, unbiased estimators which are linear combinations of order statistcs are best linear unbiased estimatos. But some setimators which are linear combiantions of order statistics may not be the best unbiased estimators.

Although all estimators of linear combinations of order statistics are less efficient than the best unbiased estimator, there are manu advantages in the use of such linear estimators.

A better linear estimator, based on the order statistics, may be obtained by the generalized least-squares estimation procedure.

In many cases, unbiased estimators which are linear combinations of order statistcs are best linear unbiased estimatos. But some setimators which are linear combiantions of order statistics may not be the best unbiased estimators.

Although all estimators of linear combinations of order statistics are less efficient than the best unbiased estimator, there are manu advantages in the use of such linear estimators.

A better linear estimator, based on the order statistics, may be obtained by the generalized least-squares estimation procedure.