屠星劍,訾雪旻
基于馬氏距離與絕對(duì)值距離的MEWMA數(shù)據(jù)實(shí)際重要性比較
屠星劍,訾雪旻
(天津職業(yè)技術(shù)師范大學(xué) 理學(xué)院,天津 300222)
基于MEWMA數(shù)據(jù)實(shí)際重要性,研究控制圖在馬氏距離和絕對(duì)值距離下的不同表現(xiàn).通過(guò)數(shù)值模擬得到了不同參數(shù)情況下的控制限和平均運(yùn)行長(zhǎng)度.結(jié)果表明,隨著維度的增大,絕對(duì)值距離比馬氏距離具有更好的監(jiān)控效果.
實(shí)際重要性;馬氏距離;絕對(duì)值距離;MEWMA
隨著監(jiān)控環(huán)境日益復(fù)雜,傳統(tǒng)的假設(shè)檢驗(yàn)并不能滿足人們的需求,帶有實(shí)際重要性的假設(shè)檢驗(yàn)成為了當(dāng)下的熱點(diǎn)問(wèn)題之一.帶有實(shí)際重要性的假設(shè)檢驗(yàn)?zāi)軌虮苊獗O(jiān)控過(guò)程中過(guò)高的敏感性,使控制圖既具有統(tǒng)計(jì)意義,又具有實(shí)際意義.Dette[1]等考慮了帶有實(shí)際重要性的假設(shè)檢驗(yàn),研究了變點(diǎn)問(wèn)題的相關(guān)變化;Woodall[2]等提出了三區(qū)域法,考慮了CUSUM控制圖的實(shí)際重要性問(wèn)題.本文在具有實(shí)際重要性的MEWMA在線控制圖的基礎(chǔ)上,考慮2種不同范數(shù)下平均運(yùn)行長(zhǎng)度(ARL)的不同表現(xiàn).
在統(tǒng)計(jì)過(guò)程控制中,多數(shù)文獻(xiàn)考慮的是經(jīng)典假設(shè)檢驗(yàn),其形式為
在統(tǒng)計(jì)學(xué)中,常用的距離有絕對(duì)值距離、馬氏距離、歐氏距離和閔可夫斯基距離等,本文將絕對(duì)值距離和馬氏距離分別與帶有實(shí)際重要性的MEWMA控制圖結(jié)合,比較在不同距離下ARL的表現(xiàn).
馬氏距離是由印度統(tǒng)計(jì)學(xué)家馬哈拉諾比斯(P.C.Mahalanobis)提出的,通過(guò)采用協(xié)方差來(lái)計(jì)算2個(gè)點(diǎn)之間的距離,能夠有效地計(jì)算2個(gè)未知樣本集之間的相似性[4].基于馬氏距離的度量學(xué)習(xí)方法是人臉識(shí)別和親屬關(guān)系識(shí)別的有效方法.
表1 MEWMA控制圖的控制限
表2 時(shí)MEWMA控制圖的ARL
表3 時(shí)MEWMA控制圖的ARL
表4 時(shí)MEWMA控制圖的ARL
本文在MEWMA控制圖和數(shù)據(jù)實(shí)際重要性結(jié)合的基礎(chǔ)上,研究了針對(duì)不同維度和不同容許度時(shí)OCARL的表現(xiàn).結(jié)果表明,隨著維度變大,絕對(duì)值距離比馬氏距離具有更好的監(jiān)控效果.今后將研究回歸模型下MEWMA控制圖的相關(guān)表現(xiàn).
[1] Dette H,Wied D.Detecting relevant changes in time series models[J].SSRN Electronic Journal,2014,78(2):371-394
[2] Woodall W H,F(xiàn)altin F W.Rethinking control chart design and evaluation[J].Quality Engineering,2019,31(4):596-605
[3] Lowry C A,Woodall W H,Champ C W.Multivariate Exponentially Weighted Moving Average Control Chart[J].Technometrics,1992,34(1):46-53
[4] 何曉群.多元統(tǒng)計(jì)分析[M].2版.北京:中國(guó)人民大學(xué)出版社,2008
[5] Roberts S W.Control Chart Tests Based on Geometric Moving Averages[J].Technometrics,1959(3):239-250
[6] Hunter J S.The Exponentially Weighted Moving Average[J].Journal of Quality Technology,1986,18(4):203-210
[7] Lucas J M,Saccucci M S.Exponentially Weighted Moving Average Control Schemes:Properties and Enhancements[J].Technometrics,1990,32(1):1-12
[8] Montgomery D C.Introduction to Statistical Process Control[M].7thed.Hoboken:John Wiley & Sons,2013
[9] Prabhu S S,Runger G C.Designing a Multivariate EWMA Control Chart[J].Journal of Quality Technology,1997,29(1): 8-15
[10] 王兆軍,鞏震,鄒長(zhǎng)亮.ARL計(jì)算方法綜述[J].?dāng)?shù)理統(tǒng)計(jì)與管理,2016(3):91-121
Comparison of practical importance of MEWMA data based on Mahalanobis distance and absolute value distance
TU Xingjian,ZI Xuemin
(School of Science,Tianjin University of Technology and Education,Tianjin 300222,China)
Based on the practical importance of MEWMA data,the different performance of control charts under Mahalanobis distance and absolute value distance was researched.The control limit and the average run length under different parameters are obtained by numerical simulation.The results show that the absolute value distance has better monitoring effect than Mahalanobis distance as the dimension increases.
practical importance;Mahalanobis distance;absolute value distance;MEWMA
O213.1文獻(xiàn)標(biāo)識(shí)碼:Adoi:10.3969/j.issn.1007-9831.2020.10.005
1007-9831(2020)10-0018-05
2020-06-10
國(guó)家自然科學(xué)基金項(xiàng)目(11771332)
屠星劍(1996-),女,內(nèi)蒙古通遼人,在讀碩士研究生,從事統(tǒng)計(jì)過(guò)程控制研究.E-mail:961039533@qq.com