Exact inequalities for sums of asymmetric random

Exact inequalities for sums of asymmetric random

ID:40378049

大小:475.49 KB

页数:31页

时间:2019-08-01

Exact inequalities for sums of asymmetric random_第1页
Exact inequalities for sums of asymmetric random_第2页
Exact inequalities for sums of asymmetric random_第3页
Exact inequalities for sums of asymmetric random_第4页
Exact inequalities for sums of asymmetric random_第5页
资源描述:

《Exact inequalities for sums of asymmetric random》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、Probab.TheoryRelat.Fields(2007)139:605635DOI10.1007/s00440-007-0055-4Exactinequalitiesforsumsofasymmetricrandomvariables,withapplicationsIosifPinelisReceived:25May2006/Revised:26December2006/Publishedonline:7February2007©Springer-Verlag2007AbstractLetB

2、S1,...,BSnbeindependentidenticallydistributedrandomvariableseachhavingthestandardizedBernoullidistributionwithparame-terp∈(0,1).Letm2221∗(p):=(1+p+2p)/(2p−p+4p)if0

3、ex.Thenitisprovedthatforallnonnegativenumbersc1,...,cnonehastheinequalityBS+···+c(m)Ef(c11nBSn)Efs(BS1+···+BSn),1wheres(m):=1nc2m2m.Thelowerboundmni=1i∗(p)onmisexactforeachp∈(0,1).Moreover,Ef(c1BS1+···+cnBSn)isSchur-concavein(c2m,...,c2nm).1Anumb

4、erofcorollariesareobtained,includingupperboundsongener-alizedmomentsandtailprobabilitiesof(super)martingaleswithdifferencesofboundedasymmetry,andalsoupperboundsonthemaximalfunctionofsuch(super)martingales.Applicationstogeneralizedself-normalizedsumsand

5、t-statisticsaregiven.Keywords(super)martingales·Probabilityinequalities·Generalizedmoments·Self-normalizedsums·t-statisticMathematicsSubjectClassification(2000)Primary:60E15·60G50·60G42·60G48·62F03·62F25·62G10·60G15;Secondary:60E05·62E10·62G35I.Pinelis(

6、B)DepartmentofMathematicalSciences,MichiganTechnologicalUniversity,Houghton,MI49931,USAe-mail:ipinelis@mtu.edu606I.Pinelis1IntroductionExponentialupperbounds,sayoftheforme−x2/(2B2),onthetailsprobabilitiesofsumsofindependentrandomvariables(r.v.s)or,more

7、generally,martingaleshavehadmanyapplicationsinprobabilityandstatistics.Intheapplications,theratioB/xisusuallysmall.Suchexponentialboundsarebasedonupperboundsonthecorrespond-ingexponentialmoments.However,incomparisonwiththeideal,normalestimate1−Φ(x/B)∼√

8、Be−x2/(2B2)ofthetailprobability(whereΦ(u):=x2πP(Zu),withZ∼N(0,1)),theexponentialupperbounde−x2/(2B2)missestheusuallysmallfactorB/x.Theapparentcauseofthisdeficiencyisthattheclassofexponentialmomentfunctionsistoosmall(andsoistheclassofth

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。