有偏估计与LS估计的比较与选择_归庆明.pdf

32 　 1 　　　 Vol.32, N o.1 　200 3 2 A CTA G EO DA ET ICA e t CART OG RA PHICA SI NI CA Feb., 200 3 　　:1001-1595 (200 3)01-0026-05 :P207 　　　　:A LS 1, 2 2 1 归庆明 , 李国重 , 欧吉坤 (1. , 4 30077;2. , 4 50052 ) Comparison and Selection between Biased Estimator and LS Estimator 1, 2 2 1 GU I Q ng-m ng , LI Guo-zhong , OU J -kuen (1.Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 4 30077, China;2.Institute of Surveying and Mapping, Inf ormation Engineering University, Zhengzhou 4 50052, China) Abstract:T he problem of select on be tw een b ased est mator and LS est mato r n G au ss-M arkov model s stud ed by u s ng the hypothes s test ng approach.F rstly, the compar sons betw een the tw o most mportant b ased est mators, o rd nary r dge est mator and pr nc pal components est mator, and LS est mator are conducted by u s ng the cr ter on of mean squared error;and the cond t ons to show the super or ty of each of these tw o est mators over the LS est ma- tor have been obta ned.T hen, the tests have been suggested to ver fy w hether or not these cond t o ns hold n g ven s tuat ons by us ng the stat st cal method.F nally, the computat onal results demonstrate that f the null hypothes s s accepted w th a s gn f cance level, w e have to bel ev e the reasonab l ty of b ased est mator nstead of LS est mato r and w e can th nk f w e u se the b ased est mato r t w ll mpro ve LS est mator mo re eff ect vely .O n the contrary, f the null hy pothes s s rejected, w e w ll su spect the b ased est mato rs super or ty .In th s case, us ng