李藝璇
(湖北大學數(shù)學與統(tǒng)計學學院,湖北 武漢 430062)
?
END樣本最近鄰密度估計的相合性
李藝璇
(湖北大學數(shù)學與統(tǒng)計學學院,湖北 武漢 430062)
在END樣本下研究最近鄰密度估計的相合性,給出弱相合性、強相合性、一致強相合性以及它們的收斂速度的充分條件, 同時研究失效函數(shù)估計的一致強相合性.
END樣本;最近鄰密度估計;相合性
為了證明本文中定理,先給出幾個引理.
引理1[4]設{X1,X2,…,Xn}是END變量.
(1) 如果f1,f2,…,fn均為非降(或非增)函數(shù),則隨機變量f1(X1),f2(X2),…,fn(Xn)也是END的.
引理2[13]設{Xi,i≥1}為END隨機變量序列,EXi=0,ai≤Xi≤bi,(i=1,2,…),則?ε>0,有
記ξ=I(Xi
因此P(supx|Fn(x)-F(x)|>ετn)≤Cn-2.由此即得結論,證畢.
引理5設X1,X2,…,Xn為END樣本,F(xiàn)(x)為連續(xù)的分布函數(shù),則
引理5的證明取τn=n-1/2(logn)1/2loglogn,由引理4即得,證畢.
引理6設X1,X2,…,Xn為END樣本,EXi=0,|Xi|≤b,(i=1,2,…),1
推論1設f(x)在x處滿足局部Lipschitz條件且f(x)>0,若取kn=[n3/4(logn)1/4],t=2,則當n→∞時,有|fn(x)-f(x)|=o(qn)a.s..
推論2設f(x)在R1處滿足Lipschitz條件.若取kn=[n2/3(logn)1/3],t=2,則當n→∞時,有
定理5設定理2的條件滿足,則對任何F(c)<1的c,均有
定理6設定理4的條件滿足,則對任何F(c)<1的c,均有
在獨立的情形下fn(x)強相合的最優(yōu)速度是n-1/3,一致強相合的最優(yōu)速度是n-1/4,在這里推論1和推論2說明fn(x)的強相合速度是n-1/4,一致強相合的速度是n-1/6與獨立情形還有距離,但是這里強相合速度和一致強相合速度與文獻[13]在NA樣本t=2的特殊情形下一樣,在這里得出相合性的最優(yōu)收斂速度與t的值成正比,t越大,相合性的收斂速度越優(yōu),要想得到最優(yōu)收斂速度,就要在指數(shù)不等式中提高t的取值,因而需要更好的指數(shù)不等式.
Ax={|fn(x)-f(x)|>ε}?
由于bn(x)→0,cn(x)→0且F′(x)=f(x),所以當n→∞時,有
因此,當n充分大時,有
(1)
(2)
同理由(2)式,有
因此就有
Ax?B1x∪B2x∪B3x∪B4x
(3)
記ξi=I(Xi
(4)
同理,有
(5)
于是,由式(3)和式(5),有
由引理7和Borel-Cantelli引理,即得定理1的結論,證畢.
定理2的證明沿用定理1證明中的Ax,A1x,B1x,bn(x),cn(x)等記號.由f(x)的一致連續(xù)性可知,存在δ>0使得當|x-y|<δ時,有|f(x)-f(y)|<ε/4.由kn/n→0知,存在正整數(shù)N>1使得當n>N時,有kn/(εn)<δ.從而有
(6)
(7)
關于x一致成立,由微分中值定理,存在θ1∈(x-bn(x),x+bn(x))和θ2∈(x-cn(x),x+cn(x))使得
F(x+bn(x))-F(x-bn(x))=2bn(x)f(θ1)
(8)
F(x+cn(x))-F(x-cn(x))=2cn(x)f(θ2)
(9)
由(6)式和(7)式知,|x-θ1|<δ,|x-θ2|<δ.于是就有
|f(x)-f(θ1)|<ε/4,|f(x)-f(θ2)|<ε/4
(10)
Fn(x+bn(x))-Fn(x-bn(x))-F(x+bn(x))+F(x-bn(x))≥
(11)
(12)
ω∈A2x,則由(9)式和(10)式,有
Fn(x+cn(x))-Fn(x-cn(x))-F(x+cn(x))+F(x-cn(x))≤
(13)
因此
(14)
由(12)式和(14)式得Ax?B.由引理3得
由此結論得證,證畢.
類似定理2的證明,因而就有
Wx?W1x∪W2x
(15)
類似定理2的證明中的(11)和(12)式,可得:對ω∈W1x,有
(16)
而對ω∈W2x,有
(17)
(18)
(19)
因此有
(20)
(21)
(22)
同理由(17)式和(21)式,有
(23)
聯(lián)合(15),(22)和(23)式,得Wx?Q1x∪Q2x∪Q3x∪Q4x.重復類似(5)式的過程,由引理7有
由此即得結論,證畢.
推論1的證明若取qn=[n-1/4(logn)1/4],kn=[n3/4(logn)1/4],t=2,由條件可以驗證定理3的條件成立.
且注意對vn(x)帶有條件f(x)>εqn,則有
證畢.
推論2的證明取kn=[n2/3(logn)1/3],qn=n-1/6(logn)1/6loglogn,t=2,則滿足定理4的條件,由定理4得出結論,證畢.
(24)
[1] Loftsgaarden D O, Quesenberry C P.A nonparametric estimate of a multivariate density function[J].The Annals of Mathematical Statistics, 1965, 36(3):1049-1051.
[2] Liu L.Precise large deviations for dependent random variables with heavy tails[J].Statistics, 2009, 79:1290-1298.
[3] Joag-Dev K, Proschan F.Negative association of random variables with applications[J].Annals of Statistics, 1983, 11(1):286-295.
[4] Liu L.Necessary and sufficient conditions for moderate deviations of dependent random variables with heavy tails[J].Science in China Series A:Mathematics, 2010, 53(6):1421-1434.
[5] Chen Y, Chen A, Ng K W.The strong law of large numbers for extend negatively dependent random variables[J].Journal of Applied Probability, 2010, 47:908-922.
[6] Shen Alting.Probability inequalities for END sequence and their applications[J].Journal of Inequalities and Applications, 2011,(12):98-110.
[7] Wu Y F, Guan M.Convergence properties of the partial sums for sequences of END random variables[J].Journal of the Korean Mathematical Society, 2012, 49(6):1097-1110.
[8] Wang X J, Zheng L L, Xu C,et al.Complete consistency for the estimator of nonparametric regression models based on extended negatively dependent errors[J].Statistics:A Journal of Theoretical and Applied Statistics, 2014, Manuscript ID 888431, DOI:10.1080/02331888.2014.888431, in press.
[9] 陳希孺.最近鄰密度估計的收斂速度[J].中國科學,1981, 12:1419-1428.
[10] Boenteb G, Fraiman R.Consistency of a nonparametric estimata of a density function for dependent variables[J].Journal of Multivariate Analysis, 1988,25(1):90-99.
[11] 柴根象.平穩(wěn)序列最近鄰密度估計的相合性[J].數(shù)學學報,1989,32(3):423-432.
[12] 楊善朝.NA樣本最近鄰密度估計的相合性[J].應用數(shù)學學報,2003,26(3):385-394.
[13] Wu J F, Zhang L X.A berry-Esseen theorem for weakly negatively dependent random variables and its applications[J].Acta Mathematical Hungarica, 2006, 110(4):293-308.
(責任編輯趙燕)
Consistency of nearest neighbor estimation of density function for extended negatively dependent samples
LI Yixuan
(Faculty of Mathematics and Statistics, Hubei University, Wuhan 430062,China)
The consistency of nearest denisty estimator for extended negatively dependent samples was discussed.Some sufficient conditions for week consistency,strong consistency,uniformly strong consistency and consistent rates were given,and then the uniformly strong consistency of the hazard rate estimation was investigated in our work.
extended negatively dependent sample; nearest neighbor density estimation; consistency
2016-02-17
李藝璇(1993-),女,碩士生
1000-2375(2016)05-0396-07
O212.2
A
10.3969/j.issn.1000-2375.2016.05.003