Scholarly article on topic 'Conditional acceptability of random variables'

Conditional acceptability of random variables Academic research paper on "Mathematics"

CC BY
0
0
Share paper
Academic journal
J Inequal Appl
OECD Field of science
Keywords
{""}

Academic research paper on topic "Conditional acceptability of random variables"

Christofides et al. Journal of Inequalities and Applications (2016) 2016:149 DOI 10.1186/s13660-016-1093-1

d Journal of Inequalities and Applications

a SpringerOpen Journal

RESEARCH Open Access

^ CrossMark

Conditional acceptability of random variables

Tasos C Christofides1*, Istvan

*Correspondence: tasos@ucy.ac.cy 1 Department of Mathematics and Statistics, University of Cyprus, P.O. Box 20537, Nicosia, 1678, Cyprus Full list of author information is available at the end of the article

Fazekas2 and Milto Hadjikyriakou3

Abstract

Acceptable random variables introduced by Giuliano Antonini etal. (J. Math. Anal. Appl. 338:1188-1203,2008) form a class of dependent random variables that contains negatively dependent random variables as a particular case. The concept of acceptability has been studied by authors under various versions of the definition, such as extended acceptability or wide acceptability. In this paper, we combine the concept of acceptability with the concept of conditioning, which has been the subject of current research activity. For conditionally acceptable random variables, we provide a number of probability inequalities that can be used to obtain asymptotic results.

MSC: 60F15; 62G20

Keywords: J"-acceptable random variables; conditional complete convergence; exponential inequalities

ft Spri

ringer

1 Introduction

Let {Xn, n e N} be a sequence of random variables defined on a probability space (fi, A, P). Giuliano Antonini etal. [1] introduced the concept of acceptable random variables as follows.

Definition 1 A finite collection of random variables X1, X2,..., Xn is said to be acceptable if for any real X,

(n \ n

xJ2 xA < ]~[ E exp(XXi).

i=1 / 1=1

An infinite sequence of random variables {Xn, n e N} is acceptable if every finite subcollection is acceptable.

The class of acceptable random variables includes in a trivial way collections of independent random variables. But in most cases, acceptable random variables exhibit a form of negative dependence. In fact, as Giuliano Antonini etal. [1] point out, negatively associated random variables with a finite Laplace transform satisfy the notion of acceptability. However, acceptable random variables do not have to be negatively dependent. A classical example of acceptable random variables that are not negatively dependent can be

© 2016 Christofides et al. This article is distributed under the terms of the Creative Commons Attribution 4.0 InternationalLicense (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use,distribution, and reproduction in any medium, provided you give appropriate credit to the originalauthor(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

constructed based on problem III.l listed in the classical book of Feller [2]. Details can be found in Giuliano Antonini et al. [l], Shen et al. [3], or Sung et al. [4].

The idea of acceptability has been modified or generalized in certain directions. For example, Giuliano Antonini etal. [l] introduced the concept of m-acceptable random variables, whereas other authors provided weaker versions such as the notions of extended acceptability or wide acceptability. The following definition given by Sung et al. [4] provides a weaker version of acceptability by imposing a condition on X.

Definition 2 A finite collection of random variables Xl, X2,...,Xn is said to be acceptable if there exists S > 0 such that, for any real |X| < S,

An infinite sequence of random variables {Xn, n e N} is acceptable if every finite subcollection is acceptable.

The concept of acceptable random variables has been studied extensively by a few authors, and a number of results are available in the literature such as exponential inequalities and complete convergence results. For the interested reader, we suggest the papers of Giuliano Antonini etal. [1], Sung etal. [4], Wang etal. [5], Shen etal. [3], among others.

Further, in addition to the definition of acceptability, Choi and Baek [6] introduced the concept of extended acceptability as follows.

Definition 3 A finite collection of random variables X1, X2,..., Xn is said to be extended acceptable if there exists a constant M >0 such that, for any real X,

An infinite sequence of random variables {Xn, n e N} is extended acceptable if every finite subcollection is extended acceptable.

It is clear that acceptable random variables are extended acceptable. The following example provides a sequence of random variables that satisfies the notion of extended acceptability.

Example 4 Let {Xn, n e N} be absolutely continuous random variables with the same distributions F(x) and densities f (x) such that the finite-dimensional distributions are given by the copula

This copula can be equivalently represented as

The density of C can be obtained by the formula

c{ui,..., Ck ) = -—--—,

d Ui d u2 • • • d Uk

and it can be proven that, for |ß | < log 2,

0 < 2-exp (|ß |) = i-^ < c(ui,...,Uk) < exp (|ß|).

Furthermore, /0 ••• /0 c(ui,..., uk) dui • • • duk = 1, which shows that c(ui,..., uk)is a density function. It is known that

fx(xi,...,xn) = c(F(xi),...,F(xn))f (xi) • • • f (Xn),

which leads to

fx(xi,... ,Xn) < exp (|ß i)f (xi) • • •f (Xn).

Hence,

exp I xi

= I i ex^^ Yxfi,,.,*) dxi ••• dxn < exp (|ß|) j • • • j ex^X ^Tx)jf (xi) • • •f (xn) dxi • • • dxn

= exp (|ß/ exp(Xxt)f (xi)dxi i=i ^ n

= exp (|ß E[exp (XXi)],

proving that the random variables satisfy the definition of extended acceptability for M = exp (|£|)> 0.

Observe that {Xn, n e N} is a strictly stationary sequence that is negatively dependent for p <0 and positively dependent for ¡i >0.

For the class of extended acceptable random variables, Choi and Baek [6] provide an exponential inequality that enables the derivation of asymptotic results based on complete convergence.

A different version of acceptability, the notion of wide acceptability, is provided by Wang etal. [5].

Definition 5 Random variables {Xn, n e N} are widely acceptable for 50 > 0 if for any real 0 < X < 50, there exist positive numbersg(n), n > 1, such that

Eexpl X J2Xi I <g(n)flEexp(XXi) for all n > i.

The following example gives random variables that satisfy the definition of wide acceptability.

Example 6 Consider a sequence {Xn, n e N} of random variables with the same absolutely continuous distribution function F(x). Assume that the joint distribution of (X^,...,Xin) is given by

Fil„„in(xi,...,xn) = Y\F(xk)(l+ a¥k(!-F(x,0)(l-F(xk)) J,

k-1 ^ 1<j<k<n '

provided that | ^1<_i<j<n ay\ < l. Then it can be proven that

fx(xi,x2,...,x„) < (1 +

1<i<j<n

Xi(xi) • • /Xn (Xn

which leads to

expl Xi

< (1+ J2 aj

\ 1<i<j<n

y\e[s* ],

proving that {Xn, n e N} is a sequence of widely extended random variables with g(n) =

1 +1 £ i<i<j<n aii \.

The concept of widely acceptable random variables follows naturally from the concept of wide dependence of random variables introduced by Wang et al. [7]. Wang et al. [8] and Wang et al. [7] stated (without proof) that, for widely orthant dependent random variables, the inequality in Definition 5 is true for any X. For widely acceptable random variables, Wang et al. [5] pointed out, although did not provide the details, that one can get exponential inequalities similar to those obtained for acceptable random variables.

In this paper, we combine the concept of conditioning on a a-algebra with the concept of acceptability (in fact, wide acceptability) and define conditionally acceptable random variables. In Section 2.1, we give the basic definitions and examples and prove some classical exponential inequalities. In Section 2.2, we provide asymptotic results by utilizing the tools of Section 2.1. Finally, in Section 3, we give some concluding remarks.

2 Results and discussion

Recently, various researchers have studied extensively the concepts of conditional independence and conditional association (see, e.g., Chow and Teicher [9], Majerak et al. [10], Roussas [11], and Prakasa Rao [12]) providing conditional versions of known results such as the generalized Borel-Cantelli lemma, the generalized Kolmogorov inequality, and the generalized Hajek-Renyi inequalities. Counterexamples are available in the literature, proving that the conditional independence and conditional association are not equivalent to the corresponding unconditional concepts.

Following the notation introduced by Prakasa Rao [12], let EF(X) = E(X\F) and PF(A) = P(A\F) denote the conditional expectation and the conditional probability, respectively,

where F is a sub-a-algebra of A. Furthermore, CovF(X, Y) denotes the conditional co-variance of X and Y given F, where

CovF(X, Y) = Ef(XY) -Ef(X)Ef(Y),

whereas the conditional variance is defined as VarF (X) = CovF (X, X).

The concept of conditional negative association was introduced by Roussas [11]. Let us recall its definition since it is related to the results presented further.

Definition 7 A finite collection of random variables XL,...,Xn is said to be conditionally negatively associated given F (F-NA) if

CovF f (Xi, i e A),g(Xj, j e B)) < 0 a.s.

for any disjoint subsets A and B of 1,2,..., n and for any real-valued componentwise non-decreasing functions/,g on R|A| and R|B|, respectively, where |A| = card(A), provided that the covariance is defined. An infinite collection is conditionally negatively associated given F if every finite subcollection is F-NA.

As mentioned earlier, it can be shown that the concepts of negative association and conditional negative association are not equivalent. See, for example, Yuan etal. [13], where various of counterexamples are given.

2.1 Conditional acceptability

In this paper, we define the concept of conditional acceptability by combining the concept of conditioning on a a-algebra and the concept of acceptability. In particular, conditioning is combined with the concept of wide acceptability. We therefore give the following definition.

Definition 8 A finite family of random variables X1, X2,...,Xn is said to be F-acceptable for S >0 ifE(exp(5|Xi|)) < to for all i and, for any |X| < S, there exist positive numbers g(n), n > 1, such that

A sequence of random variables {Xn, n e N} is F-acceptable for S >0 if every finite subfamily is F-acceptable for S >0.

Remark 9 The definition of F-acceptability allows for the quantity X to be a random object. Thus, as a result, if the random variables Xl, X2,..., Xn are F-acceptable for S > 0, then

for an F-measurable random variable X such that |X| < S a.s.

Remark 10 It can be easily verified that if random variables X1,...,Xn are F-acceptable, then the random variables X1 -EF(X1), X2 -EF(X2),..., Xn -EF(Xn) are also F-acceptable, and -X1, -X2,..., -Xn are also F-acceptable.

The random variables given in the following example satisfy the definition of F-accept-ability.

Example 11 Let U = {1,2,3,4} with P({i}) = 4. Define the events

A1 = {1,2} and A2 = {2,3} and the random variables

X1= Ia1 and X2= Ia2. Let B = {1} and F = {U, B, Bc, 0}. Then

eK, a e B,

|ex + |, a e Bc,

1, a e B,

| + | eA, a e Bc,

EF [e^i+^j =

a e B,

I(e2A + e}u + 1), a e Bc

For the random variables to satisfy the definition of F-acceptability, the following inequality needs to be valid:

Ef ^eX(X1+X2)] < g(2)EF [eXX1 ]Ef [eX ]

for all X e (-5, S) andg(2) > 0. For the case where to e B, this inequality is satisfied for all X e R if g(2) is chosen to be any number such that g(2) > 1. For the case where to e Bc, the inequality is equivalent to

(3 - 2g(2))e2X + (3 - 5g(2))eX + (3 - 2g(2)) < 0.

Observe that the last inequality is satisfied for all X e R if g(2) > |. Thus, the random variables are F-acceptable for any real X e (-S, S) where S >0 andg(2) > |. Furthermore, it is worth mentioning that the random variables {X1,X2} are not F-NA.

In the case where F is chosen to be the trivial a-algebra, that is, F = {0, U}, the definition of F-acceptability reduces to the definition of unconditional wide acceptability. The converse statement cannot always be true, and this can be proven via the following counterexample, showing that the concepts of F-acceptability and acceptability are not equivalent.

Example 12 Let ft = {1,2,3,4,5,6}, and let P({i}) = 6. Define the events Ax and A2 by

Ai = {1,2,3,4} and A2 = {2,3,4,5}

and the random variables X1 and X2 by

Xi= Iai and X2= Ia2.

Let B = {6}, and let F = {ft,B,Bc, 0} be the sub-a-algebra generated by B. Yuan etal. [13] proved that {Xx, X2} are F-NA. By proposition Pi of the same paper it follows that, for all X e R, {eAXi, eXX2} are F-NA, and therefore {X1,X2} are F-acceptable for g(2) = 1. Note that

E[exp X(Xi + X2)] = e0X( 6) + e1X( 2) + e2X( 6) = 6 (1 + 2eX + 3^), E[eXXi]E[eXX2] = (e0X(2) + e1X(^ = 1 (1 + 2eX)2,

1 (. _ , _ 1,) 1 (. _ AX2

E[expX(X1 + X2)] -Efe^jEfe^2] = -(1 + 2ex + 3e2X) — (1 + 2ex)

= 6L-1 ex + i e2x.

18 9 18

For X = log 2, this difference is positive, proving that {X1,X2} do not satisfy the definition of acceptability.

Let X be a random variable, X > 0 a.s., and let e be an F-measurable random variable such that e >0 a.s. It is known that

T, N Ef (X)

PF(X > e) < —X a.s. (1)

This inequality is a conditional version of Markov's inequality and is an tool for obtaining the results of this paper.

It is well known that exponential inequalities played an important role in obtaining asymptotic results for sums of independent random variables. Classical exponential inequalities were obtained, for example, by Bernstein, Hoeffding, Kolmogorov, Fuk, and Nagaev (see the monograph of Petrov [14]). A crucial step in proving an exponential inequality is the use of an inequality like that in Definition 2. Next, we provide several exponential inequalities for F-acceptable random variables. The following Hoeffding-type inequality is obtained by Yuan and Xie [16].

Lemma 13 Assume that P(a < X < b) = 1, where a and b are F-measurable random variables such that a < b a.s. Then

Ef[exp(X(X - Ef(X)))] < exp^ 1 X2(b - a)2^ a.s.

for any F-measurable random variable X.

The result that follows is a conditional version of the well-known Hoeffding inequality (Hoeffding [15], Theorem 2). Similar results were proven by Shen etal. [3], Theorem 2.3, for acceptable random variables and by Yuan and Xie [16], Theorem 1, for conditionally linearly negatively quadrant dependent random variables. Our result improves Theorem 1 of Yuan and Xie [16].

Theorem 14 Let Xl, X2,...,Xn be F-acceptable random variables for S > 0 such that P(ai — Xi — b,) = 1, i = 1,2,..., n, where ai and bi are F-measurable random variables such thatai < bi a.s. for all i. Then for an F-measurable e with 0 < e < | £ n=i(bi - ai)2 a.s., we have

PF(|5„ -Ef(5„)| > c) < 2g(n) exp

where Sn = £"=1 Xi.

T,h(bi - ai)2

Proof Let X be an F-measurable random variable such that 0 < X < S a.s. Then by the conditional version of Markov's inequality

A = PF(Sn - Ef(Sn) > e)

= PF(exp(x(Sn - Ef(Sn))) > eXe)

< EF[exp(X(Sn - EF(Sn)))] - eXe

EF[exp(X£n=i(Xi -EF(Xi)))]

< ggn T\EF[«p(x(X -Ef(Xi)))] e i=i

g(n) n

< — 1 lexP

X2(bi - ai)2

= g (n) exp

— J2 (bi - ai)2- Xc

where the first inequality follows by applying (1), and the second because of the F-acceptability property for X e (0, S). Since the minimum of the functionf (x) = ax2 + bx is attained at x = -2a, the minimum of the above expression is attained at X = yn (b-a-)2 • Then

A < g(n) exp = g (n) exp

EtiCbi - ai)2

L(Ln=i(bi - ai)2)2

£n=i(bi - ai)2J

L En=i(bi - ai)2J

Given that -Xl, ..., -Xn are also F-acceptable, we also have that

PF [-(Sn - EF (Sn)) > c] < g(n) exp

Eti(bi - ai)2.

The desired result follows by combining the last two expressions.

The result that follows is the conditional version of Theorem 2.1 of Shen et al. [3].

Theorem 15 Let X1,...,Xn be F-acceptable random variables for S > 0. Assume that Ef (Xi) = 0 and that Ef (X2) is finite for any i. LetB2n = YÂ=1 EF (X2). Assumethat |Xi| < cBn a.s., where cisapositive F-measurable random variable. Then

PF[Sn > eBn] <

g(n) exp[-4j-(l- y)] a.s. for ec < 1 and e < SBn,

g(n) exp[-J^] a.s. for ec > 1 and B < S

for any positive F-measurable random variable e.

Proof Let t be a positive F-measurable random variable such that tcBn < 1. Note that, for k > -,

Ef{Xk) < Ef[|Xiik--X-]

< (cBn)k--EF(X-) a.s.

Ef (etX0 = 1 + Ef(Xi) + £ l-EF Xk

< 1 + - EF (X-)

t- ( )

< 1 + t- EF (X-)

< 1 + I Ef (X?)( 1+ ^

1 + - cBn +-(cBn )- +

3 n 3-^

t tcBn tcBn

1 + - cBn +-+-+ •

3 n 3-4 3-4-5

-EF (Xf)( 1+ ^

where the second inequality follows because tcBn < 1 a.s., and the third because of the fact

3 + 3 • 4 + 3 • 4-5 + — 2.

that 3 + 34 + ^4:5 + • • • < -. Hence, for |t| < S,

Ef (etSn) = EFm etX V i=1 I

< g (n)[] Ef (,

<g(n)Hexp -Ef(X-)( 1

= g(n) exp

= g(n) exp

t- ¿ef (X-)(1+^

t-B-I1 +

where the first inequality follows by the F-acceptability property. Note that since e, Bn, and c are positive F-measurable random variables, the conditional Markov inequality can be applied, and by using the above calculations we have that, for 0 < t < S,

PF(Sn > eBn) = PF(etSn > eteBn) EF [etSn ]

< g(n)

exp[ f Bn(1 + cn )]

= g(n) exp

- ,, tcB, -teBn + - Bt \ 1 + —-n 2 n\ 2

If ec — 1 a.s., then put t = e a.s. Therefore, 0 < t < S is equivalent to e < SBn a.s. Then we obtain from (2) that

PF(S- > eBn) < g(n) exp

-H i-ec 22

If ec > 1 a.s., then put t = B a.s. Then we need B < S a.s. In this case, from (2) we obtain

PF(S- > eBn) < g(n) exp

e 3 c 4c2

< g(n) exp

Theorem 16 Let X1,...,Xn be F-acceptable random variables for S > 0. Assume that EF(Xi) = 0 a.s. and that there is an F-measurable random variable b such that |Xi |< b a.s. for all i. Define Bn = £ -=1 Ef (X2). Then, for any positive F-measurable random variable e with 2 e b < S, we have

Bn+ 3 e

PF(Sn > e) < g(n) exp

2B2n + § be

PF (|Sn |>e) < 2g(n) exp

2B2 + 2 be

Proof For t > 0, by using Taylor's expansion we have that

EF [exp(tXi)] = E

jo ¡! .

= £ \ef [Xj]

= 1 + 0 + £ tjEF X]

t2a2 ~ Г2 EF[X] .

■E-¿-^ j!

¡! 1a2

!■ 2 ai

if a2 = EF [X2] is nonzero

= 1 + -¿-Fi(t)

r t2 a2

.. ~ tj-2 Ef X ]

Fi(t) = L j--a.s.

j. ¿a2 ' 2 ai

Note that on the set where EF [X2] = 0 a.s., we have the obvious upper bound Ef[exp(tXi)] < 1 a.s. since |EF[Xj]| < EF[Xf]b/'-2 a.s. Define c = f and Mn = 1 + ^^ = 1 + ce a.s. By choosing t >0 such that t < ¿we have that

Mn -1 ce

Mn ce + B2n Observe that, for j > 2,

EF[|Xi|j] < Ef[X^-2]

< a2(3cy-2

< 1 afd-2i\ a.s. - 2 < >

Therefore,

, N ~ ti-2 2 afd-2j! Fi(t) <£ -y2 ' J

j=2 j!

j. 1 a2 >' 2 ai

= £(tc)j

1 - tc

< Mn a.s.

By applying the conditional Markov inequality and the F-acceptability of the sequence {Xn, n e N} for 0 < t < S a.s. we have

PF(Sn > e) = PF (etSn > ete) Ef [etSn ]

< g(n)U EF \eXi'\

< g (n) exp

t2 B 2

-te + t-^Mn 2 n

The minimum is obtained at t0 = a.s. It can be verified that to = —^, and therefore

By, Mm C6+Bn

the condition tc <1 a.s. is satisfied. Moreover, toc = —

i. So the above calculations

show that the choice of to =

B2„M,

a.s. is valid. Furthermore, we have assumed that 0 < t < S

a.s., and for the particular choice of t0 this restriction leads to -fa- < S, which is equivalent

B\ + ce ~ Bl + b e

< S a.s.,

which is satisfied because of the assumption of the theorem. By substituting to = elB2nMn into (3) we obtain

PF(Si > e) < g(i) exp = g (i) exp = g (i) exp

(BlMi)2 l

B\Mn 1 e2

2 B2nMu

2Bi + 2 be.

Since {-Xn, n e N} is again a sequence of ^-acceptable random variables, we obtain

PF(Si < -e) = PF (-Si > e) < g(i) exp

2Bi + | be

The probability inequalities presented above were proven under the assumption of bounded random variables. The result that follows provides a probability inequality under a moment condition.

Theorem 17 Let {Xn, n e N} be a sequence ofF-acceptable random variablesfor S > 0, and let {cn, n e N} be a sequence of positive F-measurable random variables with Cn = ti ci

for all n = 1,2,____Assume that there is a positive F-measurable random variable T with

T < S a.s. such that, for |t|< T and fixed n > 1,

EfetXk < eickt2 a.s.,k =1,2,...,n. Then, for any positive F-measurable random variable e,

PF(|Si<

2g(i)e'2Ci a.s., 0< e < CiT, 2g(i)e-T2L a.s., e > CiT.

Proof The proof follows similar arguments as those presented in Petrov [14]. Let 0 < t < T a.s. Then by applying the conditional Markov inequality and the property of F-acceptability we have that

PF(Sn > e) < e-teEf [etSn ]

< £(n)e-ten Ef [etXk ]

< g(n)e-te+^ a.s.

The aim is to minimize the above bound. Observe that, for 0 < e < CnT a.s., the above bound is minimized for t = , which satisfies the condition 0 < t < T a.s. Therefore, by

letting t = Cr we have that

PF(Sn > e) < g(n)e'2Cn a.s. (5)

Now, suppose that e > CnT. Then the minimum is obtained for t = T, and the above bound becomes of the form

PF(Sn > e) < g(n)e-^ a.s. (6)

We see that Xi,X2,...,Xn satisfy condition (4)for-T < t < 0 for some positive F-mea-surable random variables T and {cn, n e N}.So -Xi, -X2,..., -Xn are F-acceptable random variables that satisfy condition (4) for 0 < t < T, and therefore by applying inequalities (5) and (6) for -Sn we have that

PF(Sn < -e) <g(n)e'2ecn a.s. if 0 < eCnT (7)

PF(Sn < -e) <g(n)e-T2e a.s. if e > CnT.

The desired result follows by combining (5) with (7) and (6) with (8). □

Remark 18 Since {Xn, n e N} is a sequence of F-acceptable random variables withg(n) = 1, where F is the trivial a-algebra, the above result is reduced to the result of Corollary 2.1 of Shen and Wu [17].

2.2 Conditional complete convergence

Complete convergence results are well known for independent random variables (see, e.g., Gut [18]). The classical results of Hsu, Robbins, Erdos, Baum, and Katz were extended to certain dependent sequences. Using the results of Section 2.1, we can show the complete convergence for the partial sum of F-acceptable random variables under various assumptions. We will need the following definition of conditional complete convergence (see Christofides and Hadjikyriakou [19] for details).

Definition 19 A sequence of random variables {Xn, n e N} is said to converge completely given F to a random variable X if

(Xi -X| > e) < to a.s.

for any F-measurable random variable e such that e >0 a.s.

The following set of sequences, which serves purposes of brevity, was first defined by Shen etal. [3]:

H = |bn : ^hbn < to for every 0< h

The theorem that follows is a complete convergence theorem for 'self-normalized' sums.

Theorem 20 LetX1, X2,... be a sequence of F-acceptable random variables for S > 0. Assume thatEF(Xi) = 0 a.s. and \Xi\< b a.s.for all i, where b is an F-measurable random variable. Assume thatg(n) < K a.s., where K is an a.s. finite F-measurable random variable. LetB"n = n=i EF (X2) and assume that {B2n, n e NjeH a.s. Then

—2 converges completely to 0 given F.

Proof By applying the result of Theorem 16 for —b- < S a.s. we have that

t P'(") < 21gn exp(-2BTSB2;

n=1 V Bn / n=1 V n +3 bBn"

= 2 £ g(n){ exp(-2+"b

n=1 N N 3

< to a.s.

because g(n) < K a.s. and {Bn, n e N}e H a.s. □

Remark 21 In Theorem 20, the condition {Bn, n e N}e H can be substituted by B2n = O(bn), where bn e H a.s. Then it can be proven that

converges completely to 0 given F.

Here bn can be a sequence of F-measurable random variables. In such case, the result is a conditional version of Theorem 3.1 of Shen etal. [3].

Next, we provide a result, which is a conditional version of Theorem 3.2 of Shen etal. [3].

Theorem 22 LetX1, X2,... be a sequence of F-acceptable random variables for S >0. Assume that \Xi\ < c a.s., where c is an F-measurable random variable such that c >0 a.s. Assume thatg(n) < K a.s., where K is an a.s.finite F-measurable random variable. Moreover, let bn e H a.s., where bn are F-measurable. Then

Sn E (Sn) converges completely to 0 given F.

Proof By applying Theorem 14 for 0 < " < ¿S^Jj^ a.s. we have that

J2pF(\Sn -EF(Sn)| > (nbn)2")

/ x ( 2nbn"

= 2£ g(ni eXP(-¿¡2

< to a.s.

because g(n) < K a.s. and bn e H a.s.

Theorem 23 LetX1, X2,... be a sequence of F-acceptable random variables for S > 0 such that all the assumptions of Theorem 17 are satisfied. If, for any e >0 a.s.,

Eg (n)

e 2Cn < to a.s.

.s. and y^g(n)e

< to a.s.,

where {bn, n e N} is a sequence of positive F-measurable random variables, then converges completely to 0 given F.

Proof By applying Theorem 17 we have

En r sn>e < 2E g (n)

b2 €2 )e +

n=1 n=1

2J2 g(n)e

< to a.s.

The theorem that follows gives a conditional exponential inequality for the partial sum of F-acceptable random variables, under a moment condition, which, in the unconditional case is a condition appearing very frequently in large deviation results (see, e.g., Nagaev [20] and Teicher [21]). It also appears as condition (3.3) of Theorem 3.3 of Shen etal. [3]. However, the bound provided here allows us to prove the complete convergence, and in the unconditional case, under assumptions different from those of Theorem 3.3 of Shen etal. [3].

Theorem 24 Let X1, X2,... be a sequence of F-acceptable random variables for S >0. Assume that Ef(Xi) = 0 and let of = EF(X2) be a.s. finite. Let Bn = £of. Assume that there exists an a.s.positive and a.s. finite F-measurable random variable H such that

\Ef (Xm) \ < y afHm-2, Vi andm > 2 a.s.

(i) If h [1 - y 2Hx+B2 ] < S' then for an a.s. positive F-measurable random variable x, we have

> x I < 2g(n) exp

~2H2 (

2Hx + BJ-A B2

(ii) Ifg(n) < K a.s. for all n, where K is a.s.finite and {Bn} e H a.s., then

— converges completely to 0 given F.

EF [exp(tXi)] = E

t X} t3X,3

1 + tXi +-+-+ •

= 1 + trEFX2) + 6EF(X3) + - • •

< 1 + y of {1+H |t| + H 2t2 + - • • ),

where the second equality follows by the fact that EF (Xi) = 0 and t is an F-measurable random variable. If |t| < H then

EF[exp(tXi)] < 1 +

2(1- H|t|)

t2of 1

2(1- H |t|)J

For F-measurable x > 0 and 0 < t < ^.we have that

P^E Xi > xj = PF (e^ Xi > etx) EF [et Hk X ]

< g (n)e-tx]\EF [etXi\

< g (n)e-txY[ exp

2(1 - Ht)

= g(n) exp I -tx +

h(t) = -tx +

2(1 - Ht) Then h(t) is minimized at

2(1 - Ht)

2Hx + B2

and substituting this value into the RHS of (9), after some algebraic manipulations, we obtain that

P^E Xi > xl < g(n) exp

-¿G/2^+B2n-fin

Since -X1,-X2,...,-Xn is also a sequence of ^-acceptable random variables, we obtain

¿X > *) < 2g(n) exp Hv

\ t=i /

Using this inequality, we can prove that

-^(y/lHx + Bl -J Bl

ZPF( ^^ x) < * E g (i) ex^ -H*U2HBl x + B*

n=1 V l / i=i L

œ ✓ p i

= 2^g(i)expjB* -_(V2Hx + l-l)*

i=l I L J J

< œ a.s.

Remark 25 In the previous theorem, it is assumed that g( n) < K a.s. for every n, where K is finite a.s. However, we may have the complete convergence without this assumption. For example, the RHS of (10) maybe finite even when g is not bounded. Similar statements can be made for Theorems 20 and 22.

3 Conclusions

In this paper, we define the class of conditionally acceptable random variables as a generalization of the class of acceptable random variables studied previously by Giuliano Antonini et al. [1], Shen et al. [3] and Sung et al. [4], among others. The idea of conditioning on a a-algebra is gaining increasing popularity with potential applications in fields such as risk theory and actuarial science. For the class of conditionally acceptable random variables, we provide useful probability inequalities, mainly of the exponential type, which can be used to establish asymptotic results and, in particular, complete convergence results. We anticipate that the results presented in this paper will serve as a basis for research activity, which will yield further theoretical results and applications.

Competing interests

The authors declare that they have no competing interests. Authors' contributions

All authors contributed equally and read and approved the final manuscript.

Author details

1 Department of Mathematics and Statistics, University of Cyprus, P.O. Box 20537, Nicosia, 1678, Cyprus. 2 Faculty of Informatics, University of Debrecen, P.O. Box 12, Debrecen, 4010, Hungary. 3School of Sciences, UCLAN Cyprus, 12-14 University Avenue, Pyla, Larnaka, 7080, Cyprus.

Acknowledgements

The authors are grateful to the two anonymous referees for their valuable comments, which led to a much improved version of the manuscript.

Received: 23 December 2015 Accepted: 25 May 2016 Published online: 06 June 2016 References

1. Giuliano Antonini, R, Kozachenko, Y, Volodin, A: Convergence of series of dependent 0-subgaussian random variables. J. Math. Anal. Appl. 338,1188-1203 (2008)

2. Feller, W: An Introduction to Probability Theory and Its Applications, vol. II, 2nd edn. Wiley, New York (1971)

3. Shen, A, Hu, S, Volodin, A, Wang, X: Some exponential inequalities for acceptable random variables and complete convergence. J. Inequal. Appl. 2011,142 (2011)

4. Sung, SH, Srisuradetchai, P, Volodin, A: A note on the exponentialinequality for a class of dependent random variables. J. Korean Stat. Soc. 40,109-114 (2011)

5. Wang, Y, Li, Y, Gao, Q: On the exponentialinequality for acceptable random variables. J. Inequal. Appl. 2011,40 (2011)

6. Choi, J-Y, Baek, J-I: Exponentialinequalities and complete convergence of extended acceptable random variables. J. Appl. Math. Inform. 31,417-424 (2013)

7. Wang, X, Xu, C, Hu, T-C, Volodin, A, Hu, S: On complete convergence for widely orthant-dependent random variables and its applications in nonparametric regression models.Test 23(3), 607-629 (2014)

8. Wang, K, Wang, Y, Gao, Q: Uniform asymptotics for the finite-time ruin probability of a new dependent risk model with constant interest rate. Methodol. Comput. Appl. Probab. 15(1), 109-124 (2013)

9. Chow, YS, Teicher, H: Probability Theory: Independence, Interchangeability, Martingales. Springer, New York (1978)

10. Majerek, D, Nowak, W, Zieba, W: Conditionalstrong law of large numbers. Int. J. Pure Appl. Math. 20,143-157 (2005)

11. Roussas, GG: On conditionalindependence, mixing and association. Stoch. Anal. Appl. 26,1274-1309 (2008)

12. Prakasa Rao, BLS: Conditionalindependence, conditionalmixing and conditionalassociation. Ann. Inst. Stat. Math. 61,441-460 (2009)

13. Yuan, D-M, An, J, Wu, X-S: Conditionallimit theorems for conditionally negatively associated random variables. Monatshefte Math. 161,449-473 (2010)

14. Petrov, VV: Limit Theorems of Probability Theory. Oxford University Press, London (1995)

15. Hoeffding, W: Probability inequalities for sums of bounded random variables. J. Am. Stat. Assoc. 58,13-30 (1963)

16. Yuan, D-M, Xie, Y: Conditionallimit theorems for conditionally linearly negative quadrant dependent random variables. Monatshefte Math. 166, 281-299 (2012)

17. Shen, A, Wu, R: Some probability inequalities for a class of random variables and their applications. J. Inequal. Appl. 2013,57(2013)

18. Gut, A: Probability: A Graduate Course. Springer, Berlin (2005)

19. Christofides, TC, Hadjikyriakou, M: Conditionaldemimartingales and related results. J. Math. Anal. Appl. 398, 380-391 (2013)

20. Nagaev, SV: Large deviations of sums of independent random variables. Ann. Probab. 7, 745-789 (1979)

21. Teicher, H: Exponentialbounds for large deviations of sums of unbounded random variables. Sankhya 46,41-53 (1984)

Submit your manuscript to a SpringerOpen journal and benefit from:

► Convenient online submission

► Rigorous peer review

► Immediate publication on acceptance

► Open access: articles freely available online

► High visibility within the field

► Retaining the copyright to your article

Submit your next manuscript at ► springeropen.com