删失回归模型的随机逼近_英文_徐瑞峰.pdf

第33 卷第4 期 中　国　科　学　技　术　大　学　学　报 Vol.33, No.4 2003 年8 月 JOURNAL OF UNIVERSITY OF SCIENCE AND TECHNOLOGY OF CHINA Aug.2003 Article ID:0253-2778(2003)04-0379-09 Distribution Approximation in C nsor d R gr ssion Mod l by Randomly W ighting Bootstrap XU Rui-feng, FANG Yi-xin (Department of Statistics and Finance, US TC, Hef ei, China) Abstract:The le st bsolute devi tion (LAD)Estim tor of regression p r meters in the censored regression model re considered .R o nd Zh o (1992)developed the r ndom w eighting method for M-estim tes in line r regression models.In this p per w e extend these results re extented to the censored regression models. K y words:censored regression model;LAD estim tor;r ndomly w eighting CLC numb r:O212 .1　　　Docum nt cod :A AMS Subj ct Classifications (2000):Prim ry 62N02, 62G09;Second ry 62E20 0　Introduction and th main r sult We consider the line r model y = x ′β +e , 　　i =1, …, n, i i 0 i w here β is n unknown p -vector of p r meters, {x } sequence of observed p -vectors, nd 0 i {e }unobserved i.i.d.errors.i [4, 5] + Pow ell n lyzed the model th t only y =y I (y ≥0)nd {x }re observed, i =1,i i i i …, n , where I (A )is the indic tor function of set A , th t is + + y i = (x ′iβ0 +ei ) 　　i =1, …, n . (1) He proposed the le st bsolute devi tions