To mitigate the ill effects and obtain more accurate and stable solution of the observation equations,the so-called truncated singular-value decomposition(TSVD)method is introduced by modifying the small(nonzero)singu...To mitigate the ill effects and obtain more accurate and stable solution of the observation equations,the so-called truncated singular-value decomposition(TSVD)method is introduced by modifying the small(nonzero)singular values of the coefficients matrix.The proposed method directly disposed the ill-conditioning observation equations,which differs considerably from the traditional normal equation method.An application explains that TSVD can mitigate the ill conditioning.展开更多
For a system of two seerningly umrelated regressions.some general results of mean square er-ror matrix comparisons are presented.A class of linear estimators and a class of two-stage estimatorsbased on a generalized u...For a system of two seerningly umrelated regressions.some general results of mean square er-ror matrix comparisons are presented.A class of linear estimators and a class of two-stage estimatorsbased on a generalized unrestricted estimate of the dispersion matrix are proposed.Some exact finitesample properties of the two-stage estimators are obtained.展开更多
基金Project Supported by the National Natural Science Foundation of China for Distinguished Yound Scholars of China(40125013)NationalNatural Science Foundation of China(40074006,40274001)Natural Science Foundation of Henan Province,China(004051300,0211051100)
文摘To mitigate the ill effects and obtain more accurate and stable solution of the observation equations,the so-called truncated singular-value decomposition(TSVD)method is introduced by modifying the small(nonzero)singular values of the coefficients matrix.The proposed method directly disposed the ill-conditioning observation equations,which differs considerably from the traditional normal equation method.An application explains that TSVD can mitigate the ill conditioning.
基金Suppported in part by Henan Natural Setence Foundatron(004051300)
文摘For a system of two seerningly umrelated regressions.some general results of mean square er-ror matrix comparisons are presented.A class of linear estimators and a class of two-stage estimatorsbased on a generalized unrestricted estimate of the dispersion matrix are proposed.Some exact finitesample properties of the two-stage estimators are obtained.
