正在加载图片...
6 扬州大学学报(自然科学版) 第4卷 d=n0-(m-1)8 (10) vara)=(-1)2-a,/n 自析法的优点是:一个样本的高和var(@)都是有确定结果的(自举法的0:和域是随m的变化而有 波动的),并且可以证明品对总体参数8的偏差不大于对8的偏差。但自析法通常不能提供日分布 的准确信息。尤其在估计多个模型参数时(例如由3个正态分布组成的混合分布,希估计户,P, ,,,听,共8个参数),每样本仅剔除1个观察值对多个估计数的影响甚徽,从而缩小了估计 数的抽样误差,导致假阳性结果,所以近年来国际上以自举法取代自析法的趋势相当明显6, 6参考文献 to the Bootstrp[M].London:Chapman and Hall,1993 2习s DatributifrtctM们London:Chapman月 [4]Hall P.The Bootstrap and Edgeworth expansion[M]New York:Springer-Verlag.1995 [5]Weir BS.Genetic data analysis I[M]Massachusetts Sinauer Associates,Inc.Publishers.1996.163~164 385 可杨树中国医学面科全书医学计学上海 上海 学技 出版社,1985 莫意株农业试 版M John Wiley &Son,Ine..1998 aJ1.B k,1956,431353360 [1 VcherPM.HaleyCKnott SA.ConfidencenerinTLmppng byootnGneies 1995,143,1013~1020 [11]Lynch M,Walsh B.Genetics and anyis of quantitative traits [M]Mamachusett Sinauer Amociates.Ine. Publishers.1998 A RESAMPLING PROCEDURE -BOOTSTRAPPING MO Hui-dong Abstract Bootstrapping repeatedly the size n.and makes the probability 1/n for each original variate in each sampling with replacement.Thi approach can be applied to examine the fundamentel properties of sample statistic d,to estimate the stan 0.In the paper. univariate,multivariateandresidualbotstrappings,eachtogetheg with an exmpe (本文责任编辑日月 万方数据6 扬州大学学报(自然科学版) 第4卷 乱一na一(”一1)口, 1 一 } (10) var(以)一(”一1)∑(鼠一订)z/nj 自析法的优点是:一个样本的以和var(以)都是有确定结果的(自举法的9:和蹦是随m的变化而有 波动的),并且可以证明以对总体参数口的偏差不大于a对口的偏差.但自析法通常不能提供a分布 的准确信息.尤其在估计多个模型参数时(例如由3个正态分布组成的混合分布,需估计户-,户z,∥,, 声:,乒s,西,一,砖,共8个参数),每样本仅剔除1个观察值对多个估计数的影响甚微,从而缩小了估计 数的抽样误差,导致假阳性结果.所以近年来国际上以自举法取代自析法的趋势相当明显o“““】. 6参考文献 [1] Efron B.The扣cknife,出e Bootstrap,and other re蛆mpliIlg p∞∞du托s[M].Philadelphia:SIAM PublkBtioIls, 1982 [2] Efron B,Tibshiran二R.Introduction to the Bootstrap[M].Lond∞I chapm扑aTld Hall,1993 [3] Maritz J s.D诂t打bution he statistical methods[M].Lond∞:chapm蚰and Hau,1994 [4] H“P.The B∞tstrap and EdgeWorth麟pansion[M].N唧York;SpriIIger-vork.1995 [5]we.r B s.Ge鹏tic data a肿ly越s I[M].Mas蹄chu辩tts:shlauer Associat岛.I眦.PubIishers,1996.163~164, 38S [6]杨树勤.中国医学百科全书:医学统计学[M].上海:上海科学技术出版社,1985 [7]莫惠栋.农业试验统计.第2舨[M].上海:上海科技出版社.1992 [8]D哺p盯N R.smith H.Apphed regre鲒ion aIlalysis[M].New York:J0hn wdey&son,Inc.,1998 [9] QuenouiIle M H.Not岱on bias in曲tim^tionD].BiometrIk,1956,43l 353~360 [10] Vi雠her P M,}£山y c s,Knott s A.confIdence intervals h QTL mpping h B∞tstrappi雌口].Genetks, 1996.143,1 013~1 020 [11] L”ch M,w也h B.(knetics and aulyds of quantit8tive tr且I协[M].Ma$啦chu鸶tt5I si肌僻r As驰cia也s.Inc. Publishe陪。1998 A RESAMPLING PROCEDURE——B007R岁rRAPPING M0 Hui—dong (Lab of Qu蛆Gen·Y4蛆zIlou Unh·hng小伽.225。09t Chi吐) Abstnct:B00tstrapphlg rep∞tedly draws random samples,each with size n,from the original铋mpk of size n,and makes the probabihty lA for each original variate in髓ch sampHng with r印1ac锄ent.This approach can be applied to examhe the fundarnental propertoes of sampk scatistic乎,幻estitIIate the 8tan￾dard error of a and to determine the 100(1一n)%confidence intervals contamiTlg populati帆parameter 口.In the paper,univarhte,muhivariate aIld residual boot8trappings,曲ch together with an exampk, were pre鸵nted.The another re鲫phng pmcedure—Jackknifhg wa8 als0 di∞u鼬d briefIy. Keywords:re髓mpⅡ“g;univariate b00tstmpping;multiva血te b∞tstrappiIlgI residual bootstrappi“g (本文责任编辑 日 月) 万方数据
<<向上翻页向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有