两多维样本模型中位置参数相等性的M_检验_英文_.docVIP

2005 ,18 (3) :404～410 3 M2 tests f or the Equal ity of Locat ion Para meter in the Model of Two Mult ivar ia te Sa mples J IN G Pi ng (景平) 1 ,L I Yo ng (李勇) 1 , YA N G Yua n (杨元) 2 ( 1 . C hi n a U ni v e rs i t y o f M i ni n g T ec h nol o g y , B ei j i n g 100083 , C h i n a ; 2 . S o ut h w es t J i ao2 t on g U ni v e rs i t y , C he n g d u 610031 , C hi n a) Abstract :In t hi s a rticle , by t he use of P P met ho d , P P t ype M2 stati stics a nd t2 stati stic a re co nsid2 ered to test t he equalit y of locatio n p ara meter in t he mo del of t wo multivariat e sa mples. We use P P t yp e boo t st rap M2 stati stic s a nd boo t st rap t2 stati stic to app ro ximate t he di st ributio ns of t he abo ve stati stics re sp ectively. The algo rit hms fo r t he t est s a re al so given. A nd so me simulatio n experiment s a re p erfo r med. Key words :Boo t st rap met ho d ; P P met ho d ; M2 st ati stic ; t2 st ati stic CLC Number :O212 . 7 AMS( 2000) Subject Cla ssif ication :62 G10 ;62 H15 Document code :A Art icle ID :100129847 (2005) 0320404207 1 . Introduct ion Ro b u st ne ss i s o ne of ve r y i mpo r t a nt co ncep t s i n t he dat a a nal ysi s. It ha s suc h a lo ng hi s2 to r y a s t he dat a a naly si s doe s ,b ut at fi r st it wa s j u st a naive i dea a nd a si mp le met ho d. The yo ung bra nc h ro bu st st ati stic s di dn’t co me i nto bei ng until H uber P J a nd Ha mp el F R ,et al . a dva nced a set of t heo rie s i n 1960’s. Zha ng J a nd L i G Y ha ve give n a review of so me new de2 velop me nt s o n ro bu st e sti matio n a nd t e st s i n rece nt yea r s[ 9 ] . A s to o ne2di me n sio nal locatio n M2 e sti matio n , it s M2 e sti mato r T ( x) ba se d o n t he sa mp le X = ( x1 , , x n ) i s sho w n by n ∑Ψ( x i - T ( x) ) = 0 , w he re Ψi s a no n2decrea si ng f unctio n wit h Ψ( - ∞) 0 Ψ( + ∞) . i = 1 n In o t he r view ,we ca n u se ∑Ψ( x i - t) to mea sure t h