Robustness of the Least Squares Estimation
Jinfu Liu · Science Technology and Engineering · 2007
On the robustness of LSE,using the methods other than Zyskind's,some new results are gained.First,the case is discussed that the design matrix is row full order and covariance matrix is non-zero,get two s.n conditions and one sufficient condition in which GLSE is equivalent to LSE.The meaning point out that the variance of GLSE is the same as the variance of LSE.In common case,presenting three sufficient conditions,the problem is solved nicely.