Maximal inequalities and a law of the. iterated logarithm for negatively. associated random fields
|
|
- Eustace Bryant
- 5 years ago
- Views:
Transcription
1 Maximal inequalities and a law of the arxiv:math/ v1 [math.pr] 17 Oct 2006 iterated logarithm for negatively associated random fields Li-Xin ZHANG Department of Mathematics, Zhejiang University, Hangzhou , China Abstract The exponential inequality ofthe maximum partial sums is a ey to establish the law of the iterated logarithm of negatively associated random variables. In the one-indexed random sequence case, such inequalities for negatively associated random variables are established by Shao (2000) by using his comparison theorem between negatively associated and independent random variables. In the multi-indexed random field case, the comparison theorem fails. The purpose of this paper is to establish the Kolmogorov exponential inequality as well a moment inequality of the maximum partial sums of a negatively associated random field via a different method. By using these inequalities, the sufficient and necessary condition for the law of the iterated logarithm of a negatively associated random field to hold is obtained. Key Words: negative association, law of the iterated logarithm, random field, Kolmogorov exponential inequality, the maximum partial sums. Abbreviated Title: LIL for NA Fields AMS 2000 subject classifications. Primary 60F15 Research supported by National Natural Science Foundation of China 1
2 1 Introduction and the law of iterated logarithm A finite family of random variables {X i ;1 i n} is said to be negatively associated if for every pair of disjoint subsets A and B of {1,2,,n}, Cov{f(X i ;i A),g(X j ;j B)} 0 (1.1) whenever f and g are coordinatewise non-decreasing and the covariance exists. An infinite family is negatively associated if every finite subfamily is negatively associated. The concept of the negative association was introduced by Alam and Saxena (1981) and Joag-Dev and Proschan (1983). As pointed out and proved by Joag-Dev and Proschan (1983), a number of well-nown multivariate distributions possess the NA property. Many investigators discuss the properties and limit theorems of negatively associated random variables. We refer to Joag-Dev and Proschan (1983) for fundamental properties, Newman (1984) for the central limit theorem, Matula (1992) for the three series theorem, Su, et al. (1997) for the moment inequality, Roussas (1996) for the Hoeffding inequality, and Shao (2000) for the Rosenthaltype maximal inequality and the Kolmogorov exponential inequality. Shao and Su (1999) established the law of the iterated logarithm for negatively associated random variables with finite variances. Theorem A Let {X i ;i 1} be a strictly stationary negatively associated sequence with EX 1 = 0, EX 2 1 < and σ2 := EX i=2 E(X 1X i ) > 0. Let S n = n i=1 X i. Then Here and in the squeal of this paper, logx = ln(x e) S n = σ a.s. (1.2) (2nloglogn) 1/2 Let {X n ;n N d } be a field of random variables, where d 2 is a positive integer, N d denotes the d-dimensional lattice of positive integers. Through this paper, for n = (n 1,,n d ) N d, = ( 1,, d ) N d and m N, we denote n = n 1 n d, n = n n d, n = ( 1 n 1,, d n d ) and m = ( 1 m,, d m). Also, n (resp. n) means i n i (resp. i n i ), i = 1,2,,d. Denote by S n = n X and 1 = (1,,1) N d. It is nown that, if {X n } is a field of i.i.d.r.v.s, then S n (2d n loglog n ) 1/2 = (EX2 1 )1/2 a.s. (1.3) 2
3 if and only if EX 1 = 0 and EX1 2logd 1 ( X 1 )/loglog( X 1 ) <, where n means n 1,,n d. When {X n ;n N d } is a negatively associated field of random variables, Zhang and Wen (2001a) and Zhang and Wang (1999) established the wea convergence, the law of large numbers and the complete convergence similar to those for fields of independent random variables. This paper is to establish the law of the iterated logarithm similar to (1.3) for a negatively associated random field. Theorem 1.1 Let d 2 be a positive integer, and {X n ;n N d } be a wealy stationary negatively associated field of identically distributed random variables satisfying EX 1 = 0 and EX 2 1log d 1 ( X 1 )/loglog( X 1 ) <. (1.4) Denote by Υ(j i) = Cov(X j,x i ) and σ 2 = j Zd Υ(j). Then S n = σ a.s. (1.5) (2d n loglog n ) 1/2 Theorem 1.2 Let d 2 be a positive integer, and {X n ;n N d } be a negatively associated field of identically distributed random variables satisfying (1.4). Then S n (2d n loglog n ) 1/2 (EX2 1) 1/2 a.s. The following theorem tells us that the condition (1.4) is necessary for the law of the iterated logarithm to hold. Theorem 1.3 Let d 1 be a positive integer, and {X n ;n N d } be a negatively associated field of identically distributed random variables. If then (1.4) holds. ( ) S n P (2d n loglog n ) 1/2 < > 0, (1.6) In showing the law of the iterated logarithm, a main step is to establish the exponential inequalities. In the case of d = 1, such exponential inequalities for negatively associated random variables are established by Shao (2000) by using a comparison theorem between negatively associated random variables and independent random variables. However, if d 2, such a comparison theorem fails for the maximum partial sums (cf., Bulinsi and 3
4 Suquet 2001). In section 2, we establish a Kolmogorov type exponential inequality as well as a moment inequality of the maximum partial sums of a negatively associated field via a different method. Theorems are proved in Section 3. 2 Moment inequalities and exponential inequalities First, we have the following moment inequalities and exponential inequalities for the partial sums. Lemma 2.1 Let p 2 and let {Y ; n} be a negatively associated field of random variables with EY = 0 and E Y p <. Then E {( Y i p 2(15p/lnp) p i n i n EY 2 i ) p/2 + E Y i p}. Lemma 2.2 Let {Y ; n} be a negatively associated field of random variables with zero means and finite second moments. Let T = i Y i and Bn 2 = n EY 2. Then for all x > 0 and a > 0, P(T n x) P ( max Y > a ) ( x 2 ) +exp n 2(ax+Bn) 2. (2.1) Lemma 2.3 Let {Y ; n} be a negatively associated field of random variables with EY = 0 and E Y 3 <. Denote by T = i Y i and Bn 2 = n EY 2. Then for all x > 0, P(T n xb n ) ( 1 Φ(x+1) ) 6B 2 n 12B 3 n i n 1 i j n E(Y i Y j ) E Y 3 (2.2) n where Φ(x) is the distribution of a standard normal variable. Proofs of Lemmas : In the case of d = 1, Lemma 2.1 is proved by Shao (2000), and Lemma 2.3 is proved by Shao and Su (1999). Also, (2.1) follows from the following inequality easily: P(T n x) P ( max Y > a ) { x ( x +exp n a a + B2 ) ( n xa a 2 ln +1) } Bn 2. The later is proved by Su, et al.(1997). Since Lemmas do not involve the partial order of the index set, so them are valid for d 2 also. In fact, when d 2, there is a 4
5 one-one map π : { : n} {1,2,, n }. By noting that {Y π 1 (i) : i = 1,, n } is a negatively associated sequence and n i=1 ( ) = n ( ), max ny = max i n Y π 1 (i), the results follow. In a same may, one can extend (1.2) of shao (2000) to the case of d 2. Lemma 2.4 Let {Y ; n} be a negatively associated field and {Y ; n} be a field of independent random variables such that for each, Y and Y have the same distribution. Then Ef( Y ) Ef( n ny ) for any convex function f on R, whenever the expectations exist. It shall be mentioned that it is impossible to find a one-one map π : { : n} {1,2,, n } such that m i=1 Y π 1 (i) = m Y for all m n. So, the inequalities for maximum partial sums can not be extended directly. For maximum partial sums, Zhang and Wen (2001a) established two moment inequalities. Proposition 2.1 (Zhang and Wen 2001a) Let p 2, and let {Y ; n} be a negatively associated field of random variables with EY = 0 and E Y p <. Suppose that {ǫ ; n} is a field of i.i.d.r.v.s with P(ǫ = ±1) = 1/2. Also, suppose that {ǫ ; n} is independent of {Y ; n}. Denote by T = i Y i, M n = max n T, T = i Y i, Mn = max n T and X p = (E X p ) 1/p. Then M n p 5 M n p + M n 1, (2.3) and there exists a constant A p,d depending on p and d such that E M n p A p,d { (E M n ) p + ( n EY 2 ) p/2 + E Y p}. (2.4) n Proposition 2.2 (Zhang and Wen 2001a) Let {Y ; n} be a strictly stationary negatively associated field of random variables with EY 1 = 0 and 0 < EY 2 1 <. Denote by T = i Y i. Then there exists a constant K, depending only on d, such that n 1 Emax n T2 KEY 1. 2 (2.5) 5
6 The next theorem gives the estimate of EM n in (2.4) for a non-stationary negatively associated field. Theorem 2.1 Let {Y ; n} be a negatively associated field of random variables with EY = 0 and EY 2 <. Denote by T = i Y i, M n = max n T. Then there is a constant C d depending only on d such that EM n C d ney 2. (2.6) Proof. We will prove(2.6) byinductionon d. We usetheargument of Utev andpeligrad (2003). For each n, define a n = sup Y Emax m n m Y / n EY 2 1/2 (2.7) where the supremum is taen over all fields Y := {Y i } of square integrable centered negatively associated random variables. Fix such a random field {Y i } and in addition without loss of generality assume that EY 2 = 1. n Let M be a positive integer that will be specified later. Let f(x) = (( M 1/2 ) x) M 1/2. For n define: ξ = f(y ) Ef(Y ) and η = Y ξ. Then both {ξ n } and {η n } are negatively associated fields. Since ne η 2M 1/2 n EY 2 = 2M1/2, we get Emax m n Y m Emax m n ξ m +2M1/2. To estimate Emax m ξ, we shall use a blocing procedure. m n Write i = (i 1,,i d 1 ). Tae m 0 = 0 and define the integers m recursively by m m = min m : m > m 1, Eξ 2 i,j > 1 M. j=m 1 +1i n 6
7 Note that, if we denote by l the number of integers produced by this procedure, i.e.: m 0, m 1,..., m l 1, we have 1 l 1 m =1j=m 1 +1 i,j Write S m, = m j=m 1 +1 that is S m,l = n d j=m l 1 +1 Emax m n ξ m E max max 1 lm n =: I +II. j=1 Eξ 2 i,j > l 1 M so that l M. i m ξ i,j for 1 l 1 and for convenience, m l = n d i m ξ i,j S m,j. It is obvious that +E max max 1 lm n max j ξ m 1 <j<m,t t=m 1 +1 m We evaluate the two terms in the right hand side of (2.8) separately. By the induction hypothesis and Cauchy-Schwartz inequality l l I Emax S m, C d 1 E ( m ) 2 ξ m n i,j =1 l C d 1 =1 m j=m 1 +1 =1 i n j=m 1 +1 i neξ 2 i,j C d 1M 1/2 i n Eξ 2 i K d 1M 1/2. (2.8) When d = 1, the above inequality obviously holds with C 0 = 1, since max m n does not appear. To estimate II we notice that By (2.4), we obtain By the definition of m, (II) 4 Emax l Emax =1 max j ξ m nm 1 <j<m,t t=m 1 +1 m max ξ m nm 1 <j<m,t t=m 1 +1 m A 4,d E4 max m n + m 1 t=m 1 +1 n j max j Eξ 4,t + 4 ξ m 1 <j<m,t t=m 1 +1 m m 1 t=m 1 +1 n m 1 t=m 1 +1 n Eξ 2,t 1 M, 7 4 Eξ 2,t. 2.
8 so that by using the notation (2.7) for a n and the definition of ξ we obtain m 1 t=m 1 +1 n Eξ 2,t 4 M m 1 t=m 1 +1 n Eξ 4,t 4 M 2 and E 4 max Hence we obtain max j ξ m nm 1 <j<m,t t=m 1 +1 m a4 n m 1 t=m 1 +1 n { a (II) 4 4 A n 4,d M + 4 M + 1 }. M Eξ 2,t 2 a4 n M 2. Now by (2.8) and the estimates for I and II we get Emax m n Y (A 4,d/M) 1/4 a n +(5A 4,d /M) 1/4 +C d 1 M 1/2 +2M 1/2. m Therefore, by the definition of a n, a n (A 4,d /M) 1/4 a n +(5A 4,d /M) 1/4 +C d 1 M 1/2 +2M 1/2. Letting M = [16A 4,d ]+1 yields a n 2+2(C d 1 +2)M 1/2. The proof is now completed. Combining (2.4) and (2.6) we obtain the following result on the moment inequality for the maximum partial sums. Theorem 2.2 Let p 2, and let {Y ; n} be a negatively associated field of random variables with EY = 0 and E Y p <. Then there exists a constant C p,d depending on p and d such that Emax { Y p ( Cp,d m n m n EY 2 ) p/2 + E Y p}. (2.9) n In the case of d = 1, inequality (2.9) is first obtained by Shao (2000). Before that, Su et al (1997) proved that m { Emax Y p (nmax Cp EY 2 ) p/2 +nsup E Y p}. m n n =1 8 n
9 With the same proof, one can show that (2.9) of a wealy dependent random field with lim ρ n < 1 (for definition, see Peligrad and Gut (1999)) and a ρ -mixing random field (for definition, see Zhang and Wang (1999) or Zhang and Wen (2001a)). Now, we begin to establish to following Kolmogorov type exponential inequality for maximum partial sums. Theorem 2.3 Let {Y ; n} be a negatively associated field of random variables with EY = 0 and Y b a.s for some 0 < b <. Denote by T = i Y i, M n = max n T and Bn 2 = n EY 2. Then for all x > 0, P(M n 2EM n 20x) 2 d+1 exp ( x 2 ). (2.10) 2(bx+Bn) 2 Proof. Let {ǫ ; n} is a field of i.i.d.r.v.s with P(ǫ = ±1) = 1/2. Also, assume that {ǫ ; n} is independent of {Y ; n}. Denote by T n = n ǫ Y, T n,1 = n,ǫ =1 Y and T n,2 = n,ǫ = 1 Y. First, we show that Ee Mn 2 d e 2 Mn 1 Ee 10 T n 2 d+1 e 2 Mn 1 By the Lévy inequality, we have Ee 20ǫ Y. (2.11) P{ M n x} = E Y P ǫ { M n x} 2 d E Y P ǫ { T n x} 2 d P{ T n x}, x 0, n where E Y ( ) = E[ ǫ, N d ], and E ǫ, P ǫ etc are defined similarly. Then So, from (2.3) it follows that Ee Mn = 1+EM n + 1+ M n M n 1 + = 1+ M n q=2 Ee 10 M n 2 d Ee 10 T n. EM q n q! (5 M n q + M n 1 ) q q=2 q! ( ) 2 5 Mn + M n q q 1 q=2 q! E(10 M n +2 M n 1 ) q q=2 q! E(10 M n +2 M n 1 ) q q=1 q! { = E 1+ = Ee 10 M n+2 M n 1 2 d e 2 Mn 1 Ee 10 T n. 9 (10 M n +2 M n 1 ) q } q=1 q!
10 Note that Ee 10 T n = Ee 10T n,1 10T n,2 1 2 E( e 20T n,1 +e 20T n,2 ). For fixed {ǫ ; n}, we have by Lemma 2.4 or the definition (1.1), E Y e 20T n,1 E Y e 20Y = E Y e 20ǫ Y ǫ =1 E Y e 20ǫ Y ǫ =1 ǫ =1 ǫ = 1E Y e 20ǫY = n E Y e 20ǫ Y, since E Y e 20ǫ Y e E Y (20ǫ Y ) = 1. It follows that Ee 20T n,1 = E ǫ E Y e 20T ( n,1 E ǫ E Y e 20ǫ ) Y n = E ǫ E Y e 20ǫ Y = n nee 20ǫ Y. Similarly, Ee 20T n,2 n Ee 20ǫ Y = nee 20ǫ Y. It follows that Ee 10 T n nee 20ǫ Y. Similarly, Ee 10 T n nee 20ǫ Y. (2.11) is proved. Now, from (2.11) it follows that for any x > 0 and t > 0, P(M n 2EM n 20x) e 20tx 2t Mn 1 Ee tmn 2 d+1 e 20tx nee 20tǫ Y. Since Ee 20tǫ Y { = E 1+20tǫ Y + e20tǫ Y 1 20tǫ Y (ǫ Y ) 2 (ǫ Y ) 2} 1+(e 20tb 1 20tb)b 2 EY 2 exp { (e 20tb 1 20tb)b 2 EY 2 }, it follows that P(M n 2EM n 20x) 2 d+1 e 20tx exp { (e 20tb 1 20tb)b 2 B 2 n}. 10
11 Letting 20t = 1 xb b log(1+ ) yields Bn 2 { x P(M n 2EM n 20x) 2 d+1 exp b (x b + B2 n { 2 d+1 exp x 2 } 2(bx+Bn 2). xb )log(1+ b2 Bn 2 } ) 3 Proofs of the laws of iterated logarithm We need two more lemmas. Lemma 3.1 Let d 2 and let {Y ; N d } be a negatively associated field of identically distributed random variables with EY 1 = 0 and EY 2 1 log d 1 ( Y 1 )/loglog( Y 1 ) <. Denote by T n = n Y and M n = max n T. Then M n (2d n loglog n ) 1/2 20(EY 2 1 )1/2 a.s. (3.1) Proof. Let0 < ǫ < 1beanarbitrarybutfixednumber. Letb m = ǫ 40 (EY 2 1 )1/2 (m/loglogm) 1/2, f (x) = ( b ) (x b ), g (x) = x f (x). Define First, we show that Y = f (Y ) Ef (Y ), Ŷ = g (Y ), T n = ny, M n = max T, Tn = Ŷ. n n T n E T n 0 a.s. as n. (3.2) (2 n loglog n ) 1/2 Since n E Ŷ (2 n loglog n ) 1/2 n E Y I{ Y b } (2 n loglog n ) 1/2 n C (loglog ) 1/2 EY1 2I{ Y 1 b } (2 n loglog n ) 1/2 n C 1/2 EY1 2I{ Y 1 b } n 1/2 = o(1) n 1/2 n 1/2 0 as n, (3.3) 11
12 it is enough to show that n Ŷ 0 a.s. (3.4) (2 n loglog n ) 1/2 For n = (n 1,,n d ) N d, let I(n) = { = ( 1,, d ) : 2 n i 1 i 2 n i 1,i = 1, d}. (3.4) will be true if we have and I(n)Ŷ+ (2 n loglog2 n 0 a.s. (3.5) ) 1/2 I(n)Ŷ (2 n loglog2 n 0 a.s. (3.6) ) 1/2 We show (3.5) only since (3.6) can be showed similarly. Let α m := α(m) = (2mloglogm) 1/2 and Z = (Ŷ + ) α. It is easily seen that Z = 0 if Y b, Z = Y b if b Y b +α and Z = α if Y b +α. Also, {Z ; N d } is a negatively associated field of random variables. Obviously, P(Ŷ n + Z n) P(Y n α n ) C n n (logm) d 1 P(Y 1 α m ) m=1 CEY 2 1 logd 1 ( Y 1 )/loglog( Y 1 ) <. Also by (3.3), using the notation 2 n = (2 n 1,,2 n d) for n = (n 1,,n d ) N d, one has that, I(n) EZ (2 n loglog2 n ) 1/2 2 n E Ŷ (2 n loglog2 n 0. ) 1/2 So, (3.5) is equivalent to Let I(n) (Z EZ ) (2 n loglog2 n 0 a.s. (3.7) ) 1/2 Λ(n) = I(n) EY1 2I{b < Y 1 2α }. α 2 12
13 Note that b < α if is large enough. From Lemma 2.1, it follows that for n large enough and any δ > 0, P ( I(n) (Z EZ ) δ(2 n loglog2 n ) 1/2) C E I(n) (Z EZ ) 4 (2 n loglog2 n ) 2 C(α(2 n )) 4{ ( E Z 2) 2 + I(n) C(α(2 n )) 4{ ( + I(n) I(n) +C(α(2 n )) 4{ ( + I(n) C {Λ 2 (n)+ { ( +C I(n) I(n) E Z 4} EY 2 1I{b < Y 1 α +b } ) 2 } EY1I{b 4 < Y 1 α +b } I(n) α 2 P( Y 1 α +b ) ) 2 } α 4 P( Y 1 α +b ) I(n) EY1 4I{b < Y 1 2α }} α 4 P( Y 1 α ) ) 2 + I(n) } P( Y 1 α ). Obviously, { ( n C n I(n) P( Y 1 α ) ) 2 + I(n) } P( Y 1 α ) P( Y 1 α n ) CEY 2 1 logd 1 ( Y 1 )/loglog( Y 1 ) <. Also, EY 4 1 I{b < Y 1 2α } EY 4 1 I{b n < Y 1 2α n } C C C n I(n) α 4 (logm) d 1EY 1 4I{ Y 1 2α m } m=1 m (logm) d 1 m=1 =1m= =1 =1 α 4 m n EY 4 1 I{2α 1 < Y 1 2α } α 4 m (logm) d 1 (mloglogm) 2EY 4 1 I{2α 1 < Y 1 2α } α 4 n (log) d 1 (loglog) 2(loglog)EY 2 1 I{2α 1 < Y 1 2α } CEY 2 1 log d 1 ( Y 1 )/loglog( Y 1 ) <, 13
14 and, similarly to (3.8) of Li and Wu (1989) we have Λ 2 (n) <. It follows that for arbitrary δ > 0, P ( (Z EZ ) δ(2 n loglog2 n ) 1/2) <, n I(n) n which implies (3.7) by the Borel-Cantelli lemma. Thus (3.2) holds. Now, by applying (2.10) to x = (1 + 2ǫ)(EY1 2)1/2 (2d n loglog n ) 1/2 and b = 2b n, it follows that P ( M n 2EM n 20(1+2ǫ)(EY 2 1) 1/2 (2d n loglog n ) 1/2) 2 d+1 exp{ (1+ǫ)dloglog n } 2 d+1 (log n ) (1+ǫ)d. For θ > 1 and m N d, let N m = ([θ m 1 ],,[θ m d]). It follows that P ( M Nm 2EM Nm 20(1+2ǫ)(EY1) 2 1/2 (2d N m loglog N m ) 1/2) m C m (1+ǫ)d C i d 1 i (1+ǫ)d <. m i=1 Notice EM Nm C N m 1/2 by Theorem 2.1. From the Borel-Cantelli lemma, it follows that m M Nm (2d N m loglog N m ) 1/2 20(1+2ǫ)(EY 2 1) 1/2 +2limsup m = 20(1+2ǫ)(EY 2 1) 1/2. EM Nm (2d N m loglog N m ) 1/2 So, we have limsup M n (2d n loglog n ) 1/2 max N m 1 n N m M Nm (2d N m 1 loglog N m 1 ) 1/2 20θ d/2 (1+2ǫ)(EY 2 1 )1/2 a.s. (3.8) Finally, from (3.2) and (3.8) it follows that (3.1) holds. 14
15 Lemma 3.2 Let {X ; N d } be a negatively associated field of bounded random variables with EX = 0 for all N d. Denote by S n = n X. Then for any δ > 0, max δn S n+ S n (2d n loglog n ) 1/2 80d(δ(1+2δ) d ) 1/2 sup(ex 2 )1/2 (3.9) Proof. Assume that X b a.s. with 0 < b <. Denote β = sup (EX 2 )1/2. From Theorem 2.2 it follows that Emax ( X ) 2 C n β 2. (3.10) m n m For fixed n and m n, let Y = 0 for m and Y = X otherwise, and let T = i Y i for m+δn, Bm+[δn] 2 = m+[δn] EY 2. Then for m n, B 2 m+[δn] ( m+[δn] m ) β 2 δd(1+δ) d n β 2. So by Theorem 2.3, Note that by (3.10) we have P ( max δn S m+ S m 2Emax δn S m+ S m 20x ) = P ( max T E max T 20x ) m+[δn] m+[δn] { 2 d+1 x 2 } exp 2(bx+Bm+[δn] 2 ) { 2 d+1 exp x 2 2(bx+δd(1+δ) d n β 2 ) }. (3.11) max Emax S m+ S m 2E max S 2 ( ) 1/2 E max m n δn n+[δn] n+[δn] S2 K ( n+[δn] β 2) 1/2 K(1+δ) d/2 n 1/2 β. From (3.11) it follows that for n large enough ( max P max S m+ S m 20d(1+2ǫ) ( 2δ(1+δ) d d 2 β 2 n loglog n ) ) 1/2 m n δn 2 d+1 exp { (1+ǫ)dloglog n }. 15
16 Now, let I p = {p+ : [δn]}. Then there are at most [(δ 1 +1) d ]+1 such I p s whose union covers { : n}. It follows that ( P max max S m+ S m (1+2ǫ)80d ( δ(1+2δ) d β 2 n loglog n ) ) 1/2 m n δn ( ( 1 ) ( δ +1)d +1 max P max max S m+ S m p m I p δn 40d(1+2ǫ) ( 4δ(1+2δ) d β 2 n loglog n ) ) 1/2 4( 1 ( δ +1)d max P max S m+ S m m n 2δn 20d(1+2ǫ) ( 4δ(1+2δ) d β 2 n loglog n ) 1/2 ) 4( 1 δ +1)d 2 d+1 exp ( (1+ǫ)dloglog n ). If we choose n p = ([θ p 1 ],,[θ p d]), then the sum of the above probability is finite. And then by the Borel-Cantelli lemma, p max m np max δnp S m+ S m (2d n p loglog n p ) 1/2 80d ( δ(1+2δ) d) 1/2 β a.s., which implies (3.9) easily. Now, we turn to the Proof of Theorem 1.1: We can assume σ > 0, for otherwise, we can consider the field {X n +ǫz n ;n N d } instead, where {Z n ;n N d } is a field of i.i.d. standard normal random variables and ǫ > 0 is an arbitrary number. First we show that S n σ a.s. (3.12) (2d n loglog n ) 1/2 For b > 0, let g b (x) = ( b) (x b) and h b (x) = x g b (x). Then g b (x) and h b (x) are both non-decreasing functions of x. Let X = g b (X ) Eg b (X ), X = h b (X ) Eh b (X ), S n = n X andm n = max n S. AnddefineŜ, M n similarly. ThenX = X + X and, {X ; N d } and { X ; N d } are both negatively associated fields of identically distributed random variables with EX = E X = 0 and X 2b. Then by Lemma 3.1, Ŝn (2d n loglog n ) 1/2 20(E X 2 1 )1/2 20 ( EX 2 1 I{ X 1 b} ) 1/2 0 a.s. as b. (3.13) So it suffices to show that for any ǫ > 0, S n (2d n loglog n ) 1/2 (1+ǫ)2 σ a.s. (3.14) 16
17 if b is large enogh. Let I m = { : 1 i m,i = 1,,d} and Y = i I m X ( 1)m+i. Since E( i I m X ( 1)m+i ) 2 = ES 2 1m /md σ 2 and E (Y ) 2/m X d ( 1)m+i 2EX1I{ X 2 1 b} 0 as b, i I m we can choose b and m large enough such that supey 2 (1+ǫ)2 m d σ 2. For θ > 1, let N = [θ ] =: ([θ 1 ],,[θ d]). From (2.1), it follows that P( S N m x) = P( i N Y i x) 2exp Let x = (1+ǫ) 2 σ(2dm d N loglog N ) 1/2. It follows that { x 2 } 2(2m d bx+ i N EYi 2). P ( S N m (1+ǫ) 2 σ(2dm d N loglog N ) 1/2) C C P ( S N m (1+ǫ)(2d EYi 2 loglog N ) 1/2) i N exp { (1+ǫ)dloglog N } (1+ǫ)d C i d 1 i (1+ǫ)d <. i=1 From the Borel-Cantelli lemma, it follows that = limsup So, from Lemma 3.2 it follows that limsup limsup +limsup S N m (2d N m loglog N m ) 1/2 S N m (2dm d N loglog N ) 1/2 (1+ǫ)2 σ a.s. S n (2d n loglog n ) 1/2 max N m n N +1 m S N m (2d N m loglog N m ) 1/2 max N m n N +1 m S n (2d N m loglog N m ) 1/2 S n S N m (2d N m loglog N m ) 1/2 (1+ǫ)σ +80d ( (θ 1)(1+2(θ 1)) d) 1/2 (EX 2 ) 1/2 17 a.s.
18 Letting θ 1 completes the proof of (3.14). Next we show that S n σ a.s. (2d n loglog n ) 1/2 Due to (3.13), it suffices to show that for any 0 < ǫ < 1/9, for b large enough. S n (1 9ǫ)σ a.s. (3.15) (2d n loglog n ) 1/2 For N, let m = [2 1+ǫ ], p = [ ǫ ], N = (m + p ) 4. For N d and an integer q, let m = (m 1,,m d ), p = (p 1,,p d ), N = (N 1,,N d ) and q = ( q 1,,q d ). We first show the following equality P ( S N (1 7ǫ)σ(2d N loglog N ) 1/2) =. (3.16) Set I i = I i () = {(i 1)(m +p )+1 m (i 1)(m +p )+m }, A = 1 i 4I i and B = {m : m N }\A. Then CardA = 4 m N and CardB = N CardA = o( N ). Let v i,1 = j I i X j and Clearly, S,1 = v i,1 = X j, S,2 = X j. 1 i 4 j A j B and S N = S,1 +S,2 j B EX 2 j/ N EX 2 1CardB / N 0,. From (2.1), it follows that for any ǫ > 0 P ( S,2 ǫσ(2d N loglog N ) 1/2) 2 C { exp ǫ 2 σ 2 2d N loglog N 2{bǫσ(2d N loglog N ) 1/2 + j B EX 2 j } exp { 3dloglog N } <. } 18
19 Thus in order to prove (3.16) is enough to show that P ( S,1 (1 6ǫ)σ(2d N loglog N ) 1/2) =. (3.17) Let B 2 = 1 i 4Ev2 i,1. Since and 1 i 4 E 1 i 4 E ( j I i X j ( vi,1 j I i X j ) 2 = 4 ES 2 m 4 m σ 2 N σ 2 ) 2/ N CEX 2 I{ X b} 0 as b, for b and large enough B 2 (1 ǫ)2 N σ 2. From Lemma 2.3, it follows that P ( S,1 (1 6ǫ)σ(2d N loglog N ) 1/2) P ( S,1 (1 5ǫ)(2dB 2 N loglog N ) 1/2) { 1 Φ ( 1+(1 5ǫ)(2dloglog N ) 1/2)} J,1 J,2 where Obviously, J,1 = O(1) N 1 E(v i,1 v j,1 ), 1 i j 4 J,2 = O(1) N 3/2 E v i, i 4 { ( 1 Φ 1+(1 5ǫ)(2dloglog N ) 1/2)} C (log N ) (1 4ǫ)d C (1 2ǫ)d =, and by Lemma 2.1, Also, J,2 C p N 3/2 4 ( m 3/2 (2b) 3 + m (2b) 3 ) C J,1 = O(1) N 1 2 <. Cov(X i,x j ) 1 i j N,j i p = O(1) N 1 Cov(X i,x j ) 1 i j N,j i p = O(1) Cov(X 1,X j+1 ) = O(1) Cov(X 1,X j+1 ), p j N N 1 <j N 19
20 since p i = o(n i ) for i = 1,,d. It follows that J,1 = O(1) j Cov(X 1,X j+1 ) <. Hence (3.17) is proved. Now, let C = {m : N 1 < m N }, D = {m : m N }\C and U = j C X j, V = j D X j. Then S N = U +V, CardC N and CardD = o( N ). From (2.1), it follows that for any δ > 0, P ( V δσ(2d N loglog N ) 1/2) <. (3.18) By the Borel-Cantelli lemma, we conclude that From (3.16) and (3.18), it follows that V = 0 a.s. (3.19) (2d N loglog N ) 1/2 P ( U (1 8ǫ)σ(2d N loglog N ) 1/2) =. (3.20) Note that {U ; N d } is a negatively associated filed. It follows that for any x,y and i j, P(U i x,u j y) P(U i x)p(u j y). Hence, by the generalized Borel-Cantelli lemma, (3.20) yields From (3.19) and (3.21), it follows that (3.15) holds. U (1 8ǫ)σ a.s. (3.21) (2d N loglog N ) 1/2 Proof of Theorem 1.2 is similar to that of (3.12) in which we choose m = 1 and Y = X instead. Proof of Theorem 1.3: From (1.6), it follows that there exists a constant 0 < C < such that ( ) S n P (2d n loglog n ) 1/2 < C > 0. 20
21 Then ( ) X n P < 2C > 0. (3.22) (2d n loglog n ) 1/2 Let A n = { X n 2C(2d n loglog n ) 1/2 }, A + n = {X+ n 2C(2d n loglog n )1/2 }, A n = {X n 2C(2d n loglog n ) 1/2 }. Note that { I{A + n };n Nd} and { I{A n };n Nd } are both negatively associated fields. It follows that for any m and n, It follows that { Var m m+n { Var { Var m m+n m m+n I{A + n } } } I{A n} } I{A } 2 m m+n m m+n m m+n P(A + n ), P(A n). P(A ). Hence from Lemma A.6 of Zhang and Wen (2001b), it follows that for any m ( 1 P( ma )) 2 m P(A ) 2P( m Since (3.22) implies P ( A n,i.o. ) < 1, we conclude that for some m, Then from (3.23), it follows that P( m P(A ) m A ) := β < 1. 2β (1 β 2 ) <. A ). (3.23) So, which implies P ( X 1 2C(2d loglog ) 1/2) <, m EX 2 1log d 1 ( X 1 )/loglog( X 1 ) <. (3.24) 21
22 Finally, from (3.24) and the law of the large numbers (c.f. Zhang and Wang 1999), it follows that lim n S n n EX 1 n 0 a.s. which together with (1.6) yields EX 1 = 0. And then (1.4) holds. References [1] Alam, K. and Saxena, K. M. L. (1981), Positive dependence in multivariate distributions, Comm. Statist. A 10: [2] Bulinsi, A. and Suquet, C. (2001), Normal approximation for quasi-associated random fields, Statist. Probab. Lett., 54: [3] Joag-Dev, K. and Proschan, F. (1983), Negative association of random variables, with applications, Ann Statist. 11: [4] Li, D. L and Wu, Z. Q. (1989), The law of the iterated logarithm for B-Valued random variables with multidimensional indices, Ann. Probab. 17: [5] Matula, P. (1992), A note on the almost sure convergence of sums of negatively dependent random variables. Statist. Probab. Lett. 15: [6] Móricz, F. (1983), A general moment inequality for the maximum of rectangular partial sums of multiple series, Acta. Math. Hungar. 41, [7] Newman, C. M. (1984), Asymptotic independence and limit theorems for positively and negatively dependent random variables. In: Tong, Y. L. (Ed.), Inequalities in Statistics and Probability. IMS, Hayward, CA., pp [8] Peligrad, M. and Gut, A. (1999), Almost sure results for a class of wea dependent random variables, J. Theor. Probab., 12:
23 [9] Roussas, C. G. (1996), Exponential probability inequalities with some applications, In Ferguson, T. S., Shapley, L. S. and MacQueen, J. B. (eds.), Statistics, Probability and Game Theory, IMS, Hayward, CA., pp [10] Shao, Q. M.(2000), A comparison theorem on maximum inequalities between negatively associated and independent random variables, J. Theor. Probab., 13: [11] Shao, Q. M. and Su, C. (1999), The law of the iterated logarithm for negatively associated random variables, Stochastic Processes & their Appl. 83: [12] Su, C., Zhao, L. C. and Wang, Y. B. (1997), The moment inequalities and wea convergence for negatively associated sequences, Science in China 40A: [13] Utev, S. and Peligrad, M. (2003), Maximal inequalities and an invariance principle for a class of wealy dependent random variables, J. Theor. Probab., 16: [14] Zhang, L. X. and Wang, X. Y. (1999), Convergence rates in the strong laws of asymptotically negatively associated random fields, Appl. Math.-JCU, Ser. B, 14(4): [15] Zhang, L. X. and Wen J. W. (2001b), A wea convergence for negatively associated fields, Statist. Probab. Lett., 53: [16] Zhang, L. X. and Wen, J. W. (2001a), Strong laws for sums of B-valued mixing random fields, Chinese Ann. Math., 22A: (in Chinese) Li-Xin Zhang Department of Mathematics Zhejiang University, Yuquan Campus Hangzhou , Zhejiang P.R. China stazlx@zju.edu.hz.cn 23
MOMENT CONVERGENCE RATES OF LIL FOR NEGATIVELY ASSOCIATED SEQUENCES
J. Korean Math. Soc. 47 1, No., pp. 63 75 DOI 1.4134/JKMS.1.47..63 MOMENT CONVERGENCE RATES OF LIL FOR NEGATIVELY ASSOCIATED SEQUENCES Ke-Ang Fu Li-Hua Hu Abstract. Let X n ; n 1 be a strictly stationary
More informationHan-Ying Liang, Dong-Xia Zhang, and Jong-Il Baek
J. Korean Math. Soc. 41 (2004), No. 5, pp. 883 894 CONVERGENCE OF WEIGHTED SUMS FOR DEPENDENT RANDOM VARIABLES Han-Ying Liang, Dong-Xia Zhang, and Jong-Il Baek Abstract. We discuss in this paper the strong
More informationAn almost sure central limit theorem for the weight function sequences of NA random variables
Proc. ndian Acad. Sci. (Math. Sci.) Vol. 2, No. 3, August 20, pp. 369 377. c ndian Academy of Sciences An almost sure central it theorem for the weight function sequences of NA rom variables QUNYNG WU
More informationEXPONENTIAL PROBABILITY INEQUALITY FOR LINEARLY NEGATIVE QUADRANT DEPENDENT RANDOM VARIABLES
Commun. Korean Math. Soc. (007), No. 1, pp. 137 143 EXPONENTIAL PROBABILITY INEQUALITY FOR LINEARLY NEGATIVE QUADRANT DEPENDENT RANDOM VARIABLES Mi-Hwa Ko 1, Yong-Kab Choi, and Yue-Soon Choi Reprinted
More informationTHE STRONG LAW OF LARGE NUMBERS FOR LINEAR RANDOM FIELDS GENERATED BY NEGATIVELY ASSOCIATED RANDOM VARIABLES ON Z d
ROCKY MOUNTAIN JOURNAL OF MATHEMATICS Volume 43, Number 4, 2013 THE STRONG LAW OF LARGE NUMBERS FOR LINEAR RANDOM FIELDS GENERATED BY NEGATIVELY ASSOCIATED RANDOM VARIABLES ON Z d MI-HWA KO ABSTRACT. We
More informationMi-Hwa Ko. t=1 Z t is true. j=0
Commun. Korean Math. Soc. 21 (2006), No. 4, pp. 779 786 FUNCTIONAL CENTRAL LIMIT THEOREMS FOR MULTIVARIATE LINEAR PROCESSES GENERATED BY DEPENDENT RANDOM VECTORS Mi-Hwa Ko Abstract. Let X t be an m-dimensional
More informationAlmost Sure Central Limit Theorem for Self-Normalized Partial Sums of Negatively Associated Random Variables
Filomat 3:5 (207), 43 422 DOI 0.2298/FIL70543W Published by Faculty of Sciences and Mathematics, University of Niš, Serbia Available at: http://www.pmf.ni.ac.rs/filomat Almost Sure Central Limit Theorem
More informationA Remark on Complete Convergence for Arrays of Rowwise Negatively Associated Random Variables
Proceedings of The 3rd Sino-International Symposium October, 2006 on Probability, Statistics, and Quantitative Management ICAQM/CDMS Taipei, Taiwan, ROC June 10, 2006 pp. 9-18 A Remark on Complete Convergence
More informationWittmann Type Strong Laws of Large Numbers for Blockwise m-negatively Associated Random Variables
Journal of Mathematical Research with Applications Mar., 206, Vol. 36, No. 2, pp. 239 246 DOI:0.3770/j.issn:2095-265.206.02.03 Http://jmre.dlut.edu.cn Wittmann Type Strong Laws of Large Numbers for Blockwise
More informationComplete Moment Convergence for Weighted Sums of Negatively Orthant Dependent Random Variables
Filomat 31:5 217, 1195 126 DOI 1.2298/FIL175195W Published by Faculty of Sciences and Mathematics, University of Niš, Serbia Available at: http://www.pmf.ni.ac.rs/filomat Complete Moment Convergence for
More informationThe strong law of large numbers for arrays of NA random variables
International Mathematical Forum,, 2006, no. 2, 83-9 The strong law of large numbers for arrays of NA random variables K. J. Lee, H. Y. Seo, S. Y. Kim Division of Mathematics and Informational Statistics,
More informationON THE COMPLETE CONVERGENCE FOR WEIGHTED SUMS OF DEPENDENT RANDOM VARIABLES UNDER CONDITION OF WEIGHTED INTEGRABILITY
J. Korean Math. Soc. 45 (2008), No. 4, pp. 1101 1111 ON THE COMPLETE CONVERGENCE FOR WEIGHTED SUMS OF DEPENDENT RANDOM VARIABLES UNDER CONDITION OF WEIGHTED INTEGRABILITY Jong-Il Baek, Mi-Hwa Ko, and Tae-Sung
More informationCOMPLETE QTH MOMENT CONVERGENCE OF WEIGHTED SUMS FOR ARRAYS OF ROW-WISE EXTENDED NEGATIVELY DEPENDENT RANDOM VARIABLES
Hacettepe Journal of Mathematics and Statistics Volume 43 2 204, 245 87 COMPLETE QTH MOMENT CONVERGENCE OF WEIGHTED SUMS FOR ARRAYS OF ROW-WISE EXTENDED NEGATIVELY DEPENDENT RANDOM VARIABLES M. L. Guo
More informationSelf-normalized Cramér-Type Large Deviations for Independent Random Variables
Self-normalized Cramér-Type Large Deviations for Independent Random Variables Qi-Man Shao National University of Singapore and University of Oregon qmshao@darkwing.uoregon.edu 1. Introduction Let X, X
More informationSelf-normalized laws of the iterated logarithm
Journal of Statistical and Econometric Methods, vol.3, no.3, 2014, 145-151 ISSN: 1792-6602 print), 1792-6939 online) Scienpress Ltd, 2014 Self-normalized laws of the iterated logarithm Igor Zhdanov 1 Abstract
More informationA NOTE ON THE COMPLETE MOMENT CONVERGENCE FOR ARRAYS OF B-VALUED RANDOM VARIABLES
Bull. Korean Math. Soc. 52 (205), No. 3, pp. 825 836 http://dx.doi.org/0.434/bkms.205.52.3.825 A NOTE ON THE COMPLETE MOMENT CONVERGENCE FOR ARRAYS OF B-VALUED RANDOM VARIABLES Yongfeng Wu and Mingzhu
More informationResearch Article Exponential Inequalities for Positively Associated Random Variables and Applications
Hindawi Publishing Corporation Journal of Inequalities and Applications Volume 008, Article ID 38536, 11 pages doi:10.1155/008/38536 Research Article Exponential Inequalities for Positively Associated
More information5 Birkhoff s Ergodic Theorem
5 Birkhoff s Ergodic Theorem Birkhoff s Ergodic Theorem extends the validity of Kolmogorov s strong law to the class of stationary sequences of random variables. Stationary sequences occur naturally even
More informationSTA205 Probability: Week 8 R. Wolpert
INFINITE COIN-TOSS AND THE LAWS OF LARGE NUMBERS The traditional interpretation of the probability of an event E is its asymptotic frequency: the limit as n of the fraction of n repeated, similar, and
More information4 Expectation & the Lebesgue Theorems
STA 205: Probability & Measure Theory Robert L. Wolpert 4 Expectation & the Lebesgue Theorems Let X and {X n : n N} be random variables on a probability space (Ω,F,P). If X n (ω) X(ω) for each ω Ω, does
More informationAsymptotics of random sums of heavy-tailed negatively dependent random variables with applications
Asymptotics of random sums of heavy-tailed negatively dependent random variables with applications Remigijus Leipus (with Yang Yang, Yuebao Wang, Jonas Šiaulys) CIRM, Luminy, April 26-30, 2010 1. Preliminaries
More informationASSOCIATED SEQUENCES AND DEMIMARTINGALES
ASSOCIATED SEQUENCES AND DEMIMARTINGALES B.L.S. Prakasa Rao CR RAO Advanced Institute of Mathematics, Statistics and Computer Science (AIMSCS) University of Hyderabad Campus, HYDERABAD, INDIA 500046 E-mail:blsprao@gmail.com
More informationLecture 2. We now introduce some fundamental tools in martingale theory, which are useful in controlling the fluctuation of martingales.
Lecture 2 1 Martingales We now introduce some fundamental tools in martingale theory, which are useful in controlling the fluctuation of martingales. 1.1 Doob s inequality We have the following maximal
More informationPart IA Probability. Theorems. Based on lectures by R. Weber Notes taken by Dexter Chua. Lent 2015
Part IA Probability Theorems Based on lectures by R. Weber Notes taken by Dexter Chua Lent 2015 These notes are not endorsed by the lecturers, and I have modified them (often significantly) after lectures.
More informationPRZEMYSLAW M ATULA (LUBLIN]
PROBABILITY ' AND MATHEMATICAL STATEXICS Vol. 16, Fw. 2 (19961, pp. 337-343 CONVERGENCE OF WEIGHTED AVERAGES OF ASSOCIATED RANDOM VARIABLES BY PRZEMYSLAW M ATULA (LUBLIN] Abstract. We study the almost
More informationPart IA Probability. Definitions. Based on lectures by R. Weber Notes taken by Dexter Chua. Lent 2015
Part IA Probability Definitions Based on lectures by R. Weber Notes taken by Dexter Chua Lent 2015 These notes are not endorsed by the lecturers, and I have modified them (often significantly) after lectures.
More informationProbability and Measure
Part II Year 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2018 84 Paper 4, Section II 26J Let (X, A) be a measurable space. Let T : X X be a measurable map, and µ a probability
More informationLimiting behaviour of moving average processes under ρ-mixing assumption
Note di Matematica ISSN 1123-2536, e-issn 1590-0932 Note Mat. 30 (2010) no. 1, 17 23. doi:10.1285/i15900932v30n1p17 Limiting behaviour of moving average processes under ρ-mixing assumption Pingyan Chen
More informationBrownian Motion and Conditional Probability
Math 561: Theory of Probability (Spring 2018) Week 10 Brownian Motion and Conditional Probability 10.1 Standard Brownian Motion (SBM) Brownian motion is a stochastic process with both practical and theoretical
More informationComplete Moment Convergence for Sung s Type Weighted Sums of ρ -Mixing Random Variables
Filomat 32:4 (208), 447 453 https://doi.org/0.2298/fil804447l Published by Faculty of Sciences and Mathematics, Uversity of Niš, Serbia Available at: http://www.pmf..ac.rs/filomat Complete Moment Convergence
More information4 Sums of Independent Random Variables
4 Sums of Independent Random Variables Standing Assumptions: Assume throughout this section that (,F,P) is a fixed probability space and that X 1, X 2, X 3,... are independent real-valued random variables
More informationWLLN for arrays of nonnegative random variables
WLLN for arrays of nonnegative random variables Stefan Ankirchner Thomas Kruse Mikhail Urusov November 8, 26 We provide a weak law of large numbers for arrays of nonnegative and pairwise negatively associated
More informationMean convergence theorems and weak laws of large numbers for weighted sums of random variables under a condition of weighted integrability
J. Math. Anal. Appl. 305 2005) 644 658 www.elsevier.com/locate/jmaa Mean convergence theorems and weak laws of large numbers for weighted sums of random variables under a condition of weighted integrability
More informationChp 4. Expectation and Variance
Chp 4. Expectation and Variance 1 Expectation In this chapter, we will introduce two objectives to directly reflect the properties of a random variable or vector, which are the Expectation and Variance.
More informationAn improved result in almost sure central limit theorem for self-normalized products of partial sums
Wu and Chen Journal of Inequalities and Applications 23, 23:29 http://www.ournalofinequalitiesandapplications.com/content/23//29 R E S E A R C H Open Access An improved result in almost sure central it
More informationOnline publication date: 29 April 2011
This article was downloaded by: [Volodin, Andrei] On: 30 April 2011 Access details: Access Details: [subscription number 937059003] Publisher Taylor & Francis Informa Ltd Registered in England and Wales
More informationExercises in Extreme value theory
Exercises in Extreme value theory 2016 spring semester 1. Show that L(t) = logt is a slowly varying function but t ǫ is not if ǫ 0. 2. If the random variable X has distribution F with finite variance,
More informationAsymptotic Statistics-III. Changliang Zou
Asymptotic Statistics-III Changliang Zou The multivariate central limit theorem Theorem (Multivariate CLT for iid case) Let X i be iid random p-vectors with mean µ and and covariance matrix Σ. Then n (
More informationMATH5011 Real Analysis I. Exercise 1 Suggested Solution
MATH5011 Real Analysis I Exercise 1 Suggested Solution Notations in the notes are used. (1) Show that every open set in R can be written as a countable union of mutually disjoint open intervals. Hint:
More informationLecture 3 - Expectation, inequalities and laws of large numbers
Lecture 3 - Expectation, inequalities and laws of large numbers Jan Bouda FI MU April 19, 2009 Jan Bouda (FI MU) Lecture 3 - Expectation, inequalities and laws of large numbersapril 19, 2009 1 / 67 Part
More informationA D VA N C E D P R O B A B I L - I T Y
A N D R E W T U L L O C H A D VA N C E D P R O B A B I L - I T Y T R I N I T Y C O L L E G E T H E U N I V E R S I T Y O F C A M B R I D G E Contents 1 Conditional Expectation 5 1.1 Discrete Case 6 1.2
More informationP (A G) dp G P (A G)
First homework assignment. Due at 12:15 on 22 September 2016. Homework 1. We roll two dices. X is the result of one of them and Z the sum of the results. Find E [X Z. Homework 2. Let X be a r.v.. Assume
More informationComplete q-moment Convergence of Moving Average Processes under ϕ-mixing Assumption
Journal of Mathematical Research & Exposition Jul., 211, Vol.31, No.4, pp. 687 697 DOI:1.377/j.issn:1-341X.211.4.14 Http://jmre.dlut.edu.cn Complete q-moment Convergence of Moving Average Processes under
More informationSelected Exercises on Expectations and Some Probability Inequalities
Selected Exercises on Expectations and Some Probability Inequalities # If E(X 2 ) = and E X a > 0, then P( X λa) ( λ) 2 a 2 for 0 < λ
More informationResearch Article Complete Convergence for Weighted Sums of Sequences of Negatively Dependent Random Variables
Journal of Probability and Statistics Volume 2011, Article ID 202015, 16 pages doi:10.1155/2011/202015 Research Article Complete Convergence for Weighted Sums of Sequences of Negatively Dependent Random
More informationOn complete convergence and complete moment convergence for weighted sums of ρ -mixing random variables
Chen and Sung Journal of Inequalities and Applications 2018) 2018:121 https://doi.org/10.1186/s13660-018-1710-2 RESEARCH Open Access On complete convergence and complete moment convergence for weighted
More informationFormulas for probability theory and linear models SF2941
Formulas for probability theory and linear models SF2941 These pages + Appendix 2 of Gut) are permitted as assistance at the exam. 11 maj 2008 Selected formulae of probability Bivariate probability Transforms
More informationON THE STRONG LIMIT THEOREMS FOR DOUBLE ARRAYS OF BLOCKWISE M-DEPENDENT RANDOM VARIABLES
Available at: http://publications.ictp.it IC/28/89 United Nations Educational, Scientific Cultural Organization International Atomic Energy Agency THE ABDUS SALAM INTERNATIONAL CENTRE FOR THEORETICAL PHYSICS
More informationEstimates for probabilities of independent events and infinite series
Estimates for probabilities of independent events and infinite series Jürgen Grahl and Shahar evo September 9, 06 arxiv:609.0894v [math.pr] 8 Sep 06 Abstract This paper deals with finite or infinite sequences
More informationSTAT 7032 Probability Spring Wlodek Bryc
STAT 7032 Probability Spring 2018 Wlodek Bryc Created: Friday, Jan 2, 2014 Revised for Spring 2018 Printed: January 9, 2018 File: Grad-Prob-2018.TEX Department of Mathematical Sciences, University of Cincinnati,
More informationOn Negatively Associated Random Variables 1
ISSN 1995-0802, Lobachevskii Journal of Mathematics, 2012, Vol. 33, No. 1, pp. 47 55. c Pleiades Publishing, Ltd., 2012. On Negatively Associated Random Variables 1 M. Gerasimov, V. Kruglov 1*,and A.Volodin
More informationIntroduction to Self-normalized Limit Theory
Introduction to Self-normalized Limit Theory Qi-Man Shao The Chinese University of Hong Kong E-mail: qmshao@cuhk.edu.hk Outline What is the self-normalization? Why? Classical limit theorems Self-normalized
More informationRandom Bernstein-Markov factors
Random Bernstein-Markov factors Igor Pritsker and Koushik Ramachandran October 20, 208 Abstract For a polynomial P n of degree n, Bernstein s inequality states that P n n P n for all L p norms on the unit
More informationt x 1 e t dt, and simplify the answer when possible (for example, when r is a positive even number). In particular, confirm that EX 4 = 3.
Mathematical Statistics: Homewor problems General guideline. While woring outside the classroom, use any help you want, including people, computer algebra systems, Internet, and solution manuals, but mae
More information(A n + B n + 1) A n + B n
344 Problem Hints and Solutions Solution for Problem 2.10. To calculate E(M n+1 F n ), first note that M n+1 is equal to (A n +1)/(A n +B n +1) with probability M n = A n /(A n +B n ) and M n+1 equals
More informationBrownian motion. Samy Tindel. Purdue University. Probability Theory 2 - MA 539
Brownian motion Samy Tindel Purdue University Probability Theory 2 - MA 539 Mostly taken from Brownian Motion and Stochastic Calculus by I. Karatzas and S. Shreve Samy T. Brownian motion Probability Theory
More informationLecture 22: Variance and Covariance
EE5110 : Probability Foundations for Electrical Engineers July-November 2015 Lecture 22: Variance and Covariance Lecturer: Dr. Krishna Jagannathan Scribes: R.Ravi Kiran In this lecture we will introduce
More information5 Operations on Multiple Random Variables
EE360 Random Signal analysis Chapter 5: Operations on Multiple Random Variables 5 Operations on Multiple Random Variables Expected value of a function of r.v. s Two r.v. s: ḡ = E[g(X, Y )] = g(x, y)f X,Y
More informationNormal approximation for quasi associated random fields
Normal approximation for quasi associated random fields Alexander Bulinski a,1,3 Charles Suquet b,2 a Department of Mathematics and Mechanics, Moscow State University, Moscow 119 899, Russia b Statistique
More informationA Note on the Kolmogorov-Feller Weak Law of Large Numbers
Joural of Mathematical Research with Applicatios Mar., 015, Vol. 35, No., pp. 3 8 DOI:10.3770/j.iss:095-651.015.0.013 Http://jmre.dlut.edu.c A Note o the Kolmogorov-Feller Weak Law of Large Numbers Yachu
More informationn! (k 1)!(n k)! = F (X) U(0, 1). (x, y) = n(n 1) ( F (y) F (x) ) n 2
Order statistics Ex. 4. (*. Let independent variables X,..., X n have U(0, distribution. Show that for every x (0,, we have P ( X ( < x and P ( X (n > x as n. Ex. 4.2 (**. By using induction or otherwise,
More informationReliability of Coherent Systems with Dependent Component Lifetimes
Reliability of Coherent Systems with Dependent Component Lifetimes M. Burkschat Abstract In reliability theory, coherent systems represent a classical framework for describing the structure of technical
More informationOPTIMAL TRANSPORTATION PLANS AND CONVERGENCE IN DISTRIBUTION
OPTIMAL TRANSPORTATION PLANS AND CONVERGENCE IN DISTRIBUTION J.A. Cuesta-Albertos 1, C. Matrán 2 and A. Tuero-Díaz 1 1 Departamento de Matemáticas, Estadística y Computación. Universidad de Cantabria.
More informationSupermodular ordering of Poisson arrays
Supermodular ordering of Poisson arrays Bünyamin Kızıldemir Nicolas Privault Division of Mathematical Sciences School of Physical and Mathematical Sciences Nanyang Technological University 637371 Singapore
More informationLarge deviations of empirical processes
Large deviations of empirical processes Miguel A. Arcones Abstract. We give necessary and sufficient conditions for the large deviations of empirical processes and of Banach space valued random vectors.
More informationA remark on the maximum eigenvalue for circulant matrices
IMS Collections High Dimensional Probability V: The Luminy Volume Vol 5 (009 79 84 c Institute of Mathematical Statistics, 009 DOI: 04/09-IMSCOLL5 A remark on the imum eigenvalue for circulant matrices
More informationRegression and Statistical Inference
Regression and Statistical Inference Walid Mnif wmnif@uwo.ca Department of Applied Mathematics The University of Western Ontario, London, Canada 1 Elements of Probability 2 Elements of Probability CDF&PDF
More informationIntroduction and Preliminaries
Chapter 1 Introduction and Preliminaries This chapter serves two purposes. The first purpose is to prepare the readers for the more systematic development in later chapters of methods of real analysis
More informationProbability and Measure
Chapter 4 Probability and Measure 4.1 Introduction In this chapter we will examine probability theory from the measure theoretic perspective. The realisation that measure theory is the foundation of probability
More informationAsymptotic Nonequivalence of Nonparametric Experiments When the Smoothness Index is ½
University of Pennsylvania ScholarlyCommons Statistics Papers Wharton Faculty Research 1998 Asymptotic Nonequivalence of Nonparametric Experiments When the Smoothness Index is ½ Lawrence D. Brown University
More informationLARGE DEVIATION PROBABILITIES FOR SUMS OF HEAVY-TAILED DEPENDENT RANDOM VECTORS*
LARGE EVIATION PROBABILITIES FOR SUMS OF HEAVY-TAILE EPENENT RANOM VECTORS* Adam Jakubowski Alexander V. Nagaev Alexander Zaigraev Nicholas Copernicus University Faculty of Mathematics and Computer Science
More informationOn lower and upper bounds for probabilities of unions and the Borel Cantelli lemma
arxiv:4083755v [mathpr] 6 Aug 204 On lower and upper bounds for probabilities of unions and the Borel Cantelli lemma Andrei N Frolov Dept of Mathematics and Mechanics St Petersburg State University St
More informationBahadur representations for bootstrap quantiles 1
Bahadur representations for bootstrap quantiles 1 Yijun Zuo Department of Statistics and Probability, Michigan State University East Lansing, MI 48824, USA zuo@msu.edu 1 Research partially supported by
More informationLecture 9. d N(0, 1). Now we fix n and think of a SRW on [0,1]. We take the k th step at time k n. and our increments are ± 1
Random Walks and Brownian Motion Tel Aviv University Spring 011 Lecture date: May 0, 011 Lecture 9 Instructor: Ron Peled Scribe: Jonathan Hermon In today s lecture we present the Brownian motion (BM).
More informationLimiting Distributions
Limiting Distributions We introduce the mode of convergence for a sequence of random variables, and discuss the convergence in probability and in distribution. The concept of convergence leads us to the
More informationA central limit theorem for randomly indexed m-dependent random variables
Filomat 26:4 (2012), 71 717 DOI 10.2298/FIL120471S ublished by Faculty of Sciences and Mathematics, University of Niš, Serbia Available at: http://www.pmf.ni.ac.rs/filomat A central limit theorem for randomly
More informationLecture 1 Measure concentration
CSE 29: Learning Theory Fall 2006 Lecture Measure concentration Lecturer: Sanjoy Dasgupta Scribe: Nakul Verma, Aaron Arvey, and Paul Ruvolo. Concentration of measure: examples We start with some examples
More informationSome functional (Hölderian) limit theorems and their applications (II)
Some functional (Hölderian) limit theorems and their applications (II) Alfredas Račkauskas Vilnius University Outils Statistiques et Probabilistes pour la Finance Université de Rouen June 1 5, Rouen (Rouen
More informationOn the convergence of sequences of random variables: A primer
BCAM May 2012 1 On the convergence of sequences of random variables: A primer Armand M. Makowski ECE & ISR/HyNet University of Maryland at College Park armand@isr.umd.edu BCAM May 2012 2 A sequence a :
More informationOn the Set of Limit Points of Normed Sums of Geometrically Weighted I.I.D. Bounded Random Variables
On the Set of Limit Points of Normed Sums of Geometrically Weighted I.I.D. Bounded Random Variables Deli Li 1, Yongcheng Qi, and Andrew Rosalsky 3 1 Department of Mathematical Sciences, Lakehead University,
More informationThe Codimension of the Zeros of a Stable Process in Random Scenery
The Codimension of the Zeros of a Stable Process in Random Scenery Davar Khoshnevisan The University of Utah, Department of Mathematics Salt Lake City, UT 84105 0090, U.S.A. davar@math.utah.edu http://www.math.utah.edu/~davar
More informationSpatial autoregression model:strong consistency
Statistics & Probability Letters 65 (2003 71 77 Spatial autoregression model:strong consistency B.B. Bhattacharyya a, J.-J. Ren b, G.D. Richardson b;, J. Zhang b a Department of Statistics, North Carolina
More informationEntropy and Ergodic Theory Lecture 15: A first look at concentration
Entropy and Ergodic Theory Lecture 15: A first look at concentration 1 Introduction to concentration Let X 1, X 2,... be i.i.d. R-valued RVs with common distribution µ, and suppose for simplicity that
More informationChapter 4. Measure Theory. 1. Measure Spaces
Chapter 4. Measure Theory 1. Measure Spaces Let X be a nonempty set. A collection S of subsets of X is said to be an algebra on X if S has the following properties: 1. X S; 2. if A S, then A c S; 3. if
More informationLecture 2: Review of Probability
Lecture 2: Review of Probability Zheng Tian Contents 1 Random Variables and Probability Distributions 2 1.1 Defining probabilities and random variables..................... 2 1.2 Probability distributions................................
More informationProbability Background
Probability Background Namrata Vaswani, Iowa State University August 24, 2015 Probability recap 1: EE 322 notes Quick test of concepts: Given random variables X 1, X 2,... X n. Compute the PDF of the second
More informationM17 MAT25-21 HOMEWORK 6
M17 MAT25-21 HOMEWORK 6 DUE 10:00AM WEDNESDAY SEPTEMBER 13TH 1. To Hand In Double Series. The exercises in this section will guide you to complete the proof of the following theorem: Theorem 1: Absolute
More informationChapter 6. Convergence. Probability Theory. Four different convergence concepts. Four different convergence concepts. Convergence in probability
Probability Theory Chapter 6 Convergence Four different convergence concepts Let X 1, X 2, be a sequence of (usually dependent) random variables Definition 1.1. X n converges almost surely (a.s.), or with
More informationarxiv:math/ v2 [math.pr] 16 Mar 2007
CHARACTERIZATION OF LIL BEHAVIOR IN BANACH SPACE UWE EINMAHL a, and DELI LI b, a Departement Wiskunde, Vrije Universiteit Brussel, arxiv:math/0608687v2 [math.pr] 16 Mar 2007 Pleinlaan 2, B-1050 Brussel,
More information1 Variance of a Random Variable
Indian Institute of Technology Bombay Department of Electrical Engineering Handout 14 EE 325 Probability and Random Processes Lecture Notes 9 August 28, 2014 1 Variance of a Random Variable The expectation
More informationPCMI Introduction to Random Matrix Theory Handout # REVIEW OF PROBABILITY THEORY. Chapter 1 - Events and Their Probabilities
PCMI 207 - Introduction to Random Matrix Theory Handout #2 06.27.207 REVIEW OF PROBABILITY THEORY Chapter - Events and Their Probabilities.. Events as Sets Definition (σ-field). A collection F of subsets
More informationCentral Limit Theorem for Non-stationary Markov Chains
Central Limit Theorem for Non-stationary Markov Chains Magda Peligrad University of Cincinnati April 2011 (Institute) April 2011 1 / 30 Plan of talk Markov processes with Nonhomogeneous transition probabilities
More information. Find E(V ) and var(v ).
Math 6382/6383: Probability Models and Mathematical Statistics Sample Preliminary Exam Questions 1. A person tosses a fair coin until she obtains 2 heads in a row. She then tosses a fair die the same number
More informationAN INEQUALITY FOR TAIL PROBABILITIES OF MARTINGALES WITH BOUNDED DIFFERENCES
Lithuanian Mathematical Journal, Vol. 4, No. 3, 00 AN INEQUALITY FOR TAIL PROBABILITIES OF MARTINGALES WITH BOUNDED DIFFERENCES V. Bentkus Vilnius Institute of Mathematics and Informatics, Akademijos 4,
More informationSpectral representations and ergodic theorems for stationary stochastic processes
AMS 263 Stochastic Processes (Fall 2005) Instructor: Athanasios Kottas Spectral representations and ergodic theorems for stationary stochastic processes Stationary stochastic processes Theory and methods
More informationOn the law of the iterated logarithm for Gaussian processes 1
On the law of the iterated logarithm for Gaussian processes 1 Miguel A. Arcones 2 Suggested running head: L.I.L. for Gaussian processes We present some optimal conditions for the compact law of the iterated
More informationPrecise Rates in Complete Moment Convergence for Negatively Associated Sequences
Commuicatios of the Korea Statistical Society 29, Vol. 16, No. 5, 841 849 Precise Rates i Complete Momet Covergece for Negatively Associated Sequeces Dae-Hee Ryu 1,a a Departmet of Computer Sciece, ChugWoo
More informationSTOCHASTIC GEOMETRY BIOIMAGING
CENTRE FOR STOCHASTIC GEOMETRY AND ADVANCED BIOIMAGING 2018 www.csgb.dk RESEARCH REPORT Anders Rønn-Nielsen and Eva B. Vedel Jensen Central limit theorem for mean and variogram estimators in Lévy based
More informationViscosity approximation method for m-accretive mapping and variational inequality in Banach space
An. Şt. Univ. Ovidius Constanţa Vol. 17(1), 2009, 91 104 Viscosity approximation method for m-accretive mapping and variational inequality in Banach space Zhenhua He 1, Deifei Zhang 1, Feng Gu 2 Abstract
More informationIntroduction to Statistical Learning Theory
Introduction to Statistical Learning Theory Definition Reminder: We are given m samples {(x i, y i )} m i=1 Dm and a hypothesis space H and we wish to return h H minimizing L D (h) = E[l(h(x), y)]. Problem
More information