We investigate the complete moment convergence of double-indexed weighted sums of martingale differences. Then it is easy to obtain the Marcinkiewicz-Zygmund-type strong law of large numbers of double-indexed weighted sums of martingale differences. Moreover, the convergence of double-indexed weighted sums of martingale differences is presented in mean square. On the other hand, we give the application to study the convergence of the state observers of linear-time-invariant systems and present the convergence with probability one and in mean square.
1. Introduction
Hsu and Robbins [1] introduced the concept of complete convergence; that is, a sequence of random variables {Xn,n≥1} is said to converge completely to a constant C if ∑n=1∞P(|Xn-C|≥ε)<∞ for all ε>0. By Borel-Cantelli lemma, it follows that Xn→C almost surely (a.s.). The converse is true if {Xn,n≥1} is independent. But the converse cannot always be true for the dependent case. Hsu and Robbins [1] obtained that the sequence of arithmetic means of independent and identically distributed (i.i.d.) random variables converges completely to the expected value if the variance of the summands is finite. Erdös [2] proved the converse. The result of Hsu-Robbins-Erdös is a fundamental theorem in probability theory, and it has been generalized and extended in several directions by many authors. Baum and Katz [3] gave the following generalization to establish a rate of convergence in the sense of Marcinkiewicz-Zygmund-type strong law of large numbers.
Theorem 1.
Let α>1/2, αp>1, and {Xn,n≥1} be a sequence of i.i.d. random variables. Assume that EX1=0 if α≤1. Then the following statements are equivalent:
E|X1|p<∞;
∑n=1∞nαp-2P(max1≤k≤n|∑i=1kXi|>εnα)<∞ for all ε>0.
Many authors have extended Theorem 1 to the martingale differences. For example, Yu [4] obtained the complete convergence for weighted sums of martingale differences; Ghosal and Chandra [5] gave the complete convergence of martingale arrays; Stoica [6, 7] investigated the Baum-Katz-Nagaev-type results for martingale differences and the rate of convergence in the strong law of large numbers for martingale differences; Wang et al. [8] also studied the complete convergence and complete moment convergence for martingale differences, which generalized some results of Stoica [6, 7]; Yang et al. [9] obtained the complete convergence for the moving average process of martingale differences and so forth. For other works about convergence analysis, one can refer to Gut [10], Chen et al. [11], Sung [12–14], Sung and Volodin [15], Hu et al. [16], and the references therein.
In this paper, we study the moment complete convergence of double-indexed weighted sums of martingale differences. Then it is easy to obtain the Marcinkiewicz-Zygmund-type strong law of large numbers of double-indexed weighted sums of martingale differences. Moreover, the convergence of double-indexed weighted sums of martingale differences is presented in mean square. For the details, see Theorem 5, Corollary 6, and Theorem 7 in Section 2. On the other hand, we give the applications of Corollary 6 and Theorem 7 to study the convergence of the state observers of linear-time-invariant systems and present their convergence with probability one and in mean square, respectively (see Theorems 11 and 12 in Section 3).
Recall that the sequence {Xn,n≥1} is stochastically dominated by a nonnegative random variable X if supn≥1P(|Xn|>t)≤KP(X>t) for some positive constant K and for all t≥0.
Throughout the paper, let ℱ0={∅,Ω}, 1(B) be the indicator function of set Bx+=x1(x≥0), and let K,K1,K2,… denote some positive constants not depending on n, which may be different in various places.
The following lemmas are useful for the proofs of the main results.
Lemma 2 (cf. Hall and Heyde [17, Theorem 2.11]).
If {Xi,ℱi,1≤i≤n} are martingale differences and p>0, then there exists a constant K depending only on p such that
(1)E(max1≤k≤n|∑i=1kXi|p)≤K{E(∑i=1nE(Xi2∣ℱi-1))p/2+E(max1≤i≤n|Xi|p)},mmmmmmmmmmmmmmmmmmmmmmmmn≥1.
Lemma 3 (cf. Sung [12, Lemma 2.4]).
Let {Xn,n≥1} and {Yn,n≥1} be sequences of random variables. Then for any n≥1, q>1, ε>0, and a>0, one has
(2)E(max1≤j≤n|∑i=1j(Xi+Yi)|-εa)+≤(1εq+1q-1)1aq-1E(max1≤j≤n|∑i=1jXi|q)+E(max1≤j≤n|∑i=1jYi|).
Lemma 4 (cf. Wang et al. [8, Lemma 2.2]).
Let {Xn,n≥1} be a sequence of random variables stochastically dominated by a nonnegative random variable X. Then for any n≥1, a>0, and b>0, the following two statements hold:
(3)E[|Xn|a1(|Xn|≤b)]≤K1{E[Xa1(X≤b)]+baP(X>b)},E[|Xn|a1(|Xn|>b)]≤K2E[Xa1(X>b)].
Consequently, E|Xn|a≤K3EXa. Here K1, K2, and K3 are positive constants.
2. The Convergence of Double-Indexed Weighted Sums of Martingale Differences
First, we give the complete moment convergence of double-indexed weighted sums of martingale differences.
Theorem 5.
Let α>1/2, p≥2, and {Xn,ℱn,n≥1} be martingale differences stochastically dominated by a nonnegative random variable X with EXp<∞. Let {ani,1≤i≤n,n≥1} be a triangular array of real numbers. For some q>2(αp-1)/(2α-1), we assume that E[supn≥1E(Xn2∣ℱn-1)]q/2<∞ and
(4)∑i=1n|ani|q=O(n).
Then for every ε>0,
(5)∑n=1∞nαp-2-αE(max1≤k≤n|∑i=1kaniXi|-εnα)+<∞.
Taking p=2l and α=2/p for 1≤l<2 in Theorem 5, we have the following result.
Corollary 6.
Let 1≤l<2, {Xn,ℱn,n≥1} be martingale differences stochastically dominated by a nonnegative random variable X with EX2l<∞. Let {ani,1≤i≤n,n≥1} be a triangular array of real numbers. For some q>2l/(2-l), one assumes that E[supn≥1E(Xn2∣ℱn-1)]q/2<∞ and (4) holds true. Then for every ε>0,
(6)∑n=1∞n-1/lE(max1≤k≤n|∑i=1kaniXi|-εn1/l)+<∞.
In particular, one has
(7)limn→∞1n1/l∑i=1naniXi=0,a.s.
Next, we investigate the convergence in mean square.
Theorem 7.
Let r>1/2 and {Xn,ℱn,n≥1} be martingale differences stochastically dominated by a nonnegative random variable X with EX2<∞. Let {ani,1≤i≤n,n≥1} be a triangular array of real numbers and
(8)∑i=1nani2=O(n).
Then, one has
(9)n2r-1E(1nr∑i=1naniXi)2≤K,n≥1,
where K is a positive constant.
Remark 8.
Wang et al. [8] obtained the complete convergence and complete moment convergence for nonweighted martingale differences, which generalized some results of Stoica [6, 7]. In this paper, we study the complete moment convergence of double-indexed weighted sums of martingale differences. So we extend the results of Wang et al. [8] and Stoica [6, 7] to the case of double-indexed weighted sums of martingale differences. On the other hand, we give the applications of Corollary 6 and Theorem 7 to study the convergence of the state observers of linear-time-invariant systems and present the convergence with probability one and in mean square, respectively (see Theorems 11 and 12 in Section 3).
Proof of Theorem 5.
Let Xni=Xi1(|Xi|≤nα), 1≤i≤n. It can be found that aniXi=aniXi1(|Xi|>nα)+[aniXni-aniE(Xni∣ℱi-1)]+aniE(Xni∣ℱi-1),1≤i≤n.
By Lemma 3 with a=nα, for any q>1, we obtain that
(10)∑n=1∞nαp-2-αE(max1≤k≤n|∑i=1kaniXi|-εnα)+≤K1∑n=1∞nαp-2-qα×E(max1≤k≤n|∑i=1k[aniXni-aniE(Xni∣ℱi-1)]|q)+∑n=1∞nαp-2-αE(max1≤k≤n|∑i=1kaniXi1(X|Xi|>nα)|)+∑n=1∞nαp-2-αE(max1≤k≤n|∑i=1kaniE(Xni∣ℱi-1)|):=H1+H2+H3.
For p≥2, it is easy to see that q>2(αp-1)/(2α-1)≥2. Consequently, for any 1≤s≤2, we get by Hölder’s inequality and (4) that
(11)∑i=1n|ani|s≤(∑i=1n|ani|q)s/q(∑i=1n1)1-s/q=O(n).
So, it can be checked by Markov’s inequality, Lemma 4, (11), and EXp<∞ (p≥2) that
(12)H2≤∑n=1∞nαp-2-α∑i=1n|ani|E[|Xi|1(|Xi|>nα)]≤K1∑n=1∞nαp-1-αE[X1(X>nα)]=K1∑n=1∞nαp-1-α∑m=n∞E[X1(mα<X≤(m+1)α)]=K1∑m=1∞E[X1(mα<X≤(m+1)α)]∑n=1mnαp-1-α≤K2∑m=1∞mαp-αE[X1(mα<X≤(m+1)α)]≤K2EXp<∞.
Since {Xi,ℱi,1≤i≤n} are martingale differences, by the martingale property and the proof of (12), one has that
(13)H3=∑n=1∞nαp-2-αE(max1≤k≤n|∑i=1kaniE[Xi1(|Xi|≤nα)∣ℱi-1]|)=∑n=1∞nαp-2-αE(max1≤k≤n|∑i=1kaniE[Xi1(|Xi|>nα)∣ℱi-1]|)≤K1∑n=1∞nαp-2-α∑i=1n|ani|E[|Xi|1(|Xi|>nα)]≤K2∑n=1∞nαp-1-αE[X1(X>nα)]≤K3EXp<∞.
Next, we turn to prove H1<∞ under conditions of Theorem 5. It can be seen that
(14){[aniXni-aniE(Xni∣ℱi-1)],ℱi,1≤i≤n}
are also martingale differences. So, by Markov’s inequality, (10), and Lemma 2 with p=q, it can be found that
(15)H1=K1∑n=1∞nαp-2-qαE(max1≤k≤n|∑i=1k[aniXni-aniEVVVVVVVVVvvvvvvvvvvvvvV×(Xni∣ℱi-1)]∑i=1k|q∑i=1k)≤K2∑n=1∞nαp-2-qαE(∑i=1nE{[aniXni-aniEVsssssssssssssssssssvvvvvsss×(Xni∣ℱi-1)]2∣ℱi-1}∑i=1n)q/2+K3∑n=1∞nαp-2-qα∑i=1nE|aniXni-aniE(Xni∣ℱi-1)|q=:K2H11+K3H12.
Obviously, it follows that
(16)E{[aniXni-E(aniXni∣ℱi-1)]2∣ℱi-1}=E[ani2Xi21(|Xi|≤nα)∣ℱi-1]-[E(aniXi1(|Xi|≤nα)∣ℱi-1)]2≤ani2E[Xi21(|Xi|≤nα)∣ℱi-1]≤ani2E(Xi2∣ℱi-1),a.s.,1≤i≤n.
Combining (11) with E[supi≥1E(Xi2∣ℱi-1)]q/2<∞, we obtain that
(17)H11≤∑n=1∞nαp-2-qα(∑i=1nani2)q/2E(supi≥1supi≥1E(Xi2∣ℱi-1))q/2≤K4∑n=1∞nαp-2-qα+q/2<∞,
following from the fact that q>2(αp-1)/(2α-1). Meanwhile, by Cr inequality, Lemma 4, and (4),
(18)H12≤K5∑n=1∞nαp-2-qα∑i=1n|ani|qE[|Xi|q1(|Xi|≤nα)]≤K6∑n=1∞nαp-1-qαE[Xq1(X≤nα)]+K7∑n=1∞nαp-1P(X>nα)≤K6∑n=1∞nαp-1-qαE[Xq1(X≤nα)]+K7∑n=1∞nαp-1-αE[X1(X>nα)]=:K6H11*+K7H12*.
By the conditions p≥2 and α>1/2, we have that 2(αp-1)/(2α-1)-p≥0, which implies q>p. So, we obtain by EXp<∞ that
(19)H11*=∑n=1∞nαp-1-qα∑i=1nE[Xq1((i-1)α<X≤iα)]=∑i=1∞E[Xq1((i-1)α<X≤iαX)]∑n=i∞nαp-1-qα≤K8∑i=1∞E[XpXq-p1((i-1)α<X≤iα)]iαp-qα≤K8EXp<∞.
By the proof of (12), one has that
(20)H12*=∑n=1∞nαp-1-αE[X1(X>nα)]≤K9EXp<∞.
Thus, by (15)–(20), we have that H1<∞. So, it completes the proof of (5).
Proof of Corollary 6.
If p=2l and α=2/p, then one has αp=2. So as an application of Theorem 5, one gets (6) immediately. On the other hand, it can be seen that
(21)∑n=1∞nαp-2-αE(max1≤k≤n|∑i=1kaniXi|-εnα)+=∑n=1∞nαp-2-α∫0∞P(max1≤k≤n|∑i=1kaniXi|-εnα>t)dt≥∑n=1∞nαp-2-α∫0εnαP(max1≤k≤n|∑i=1kaniXi|-εnα>t)dt≥ε∑n=1∞nαp-2P(max1≤k≤n|∑i=1kaniXi|>2εnα).
So by (5) and (21) with αp=2, we have for every ε>0 that
(22)∑n=1∞P(max1≤k≤n|∑i=1kaniXi|>εn1/l)<∞.
It follows from Borel-Cantelli lemma that
(23)limn→∞1n1/l∑i=1naniXi=0,a.s.
So, (7) holds.
Proof of Theorem 7.
Since {aniXi,ℱi,1≤i≤n} are martingale differences, it can be found by Lemmas 2 and 4 and (8) that
(24)E(1nr∑i=1naniXi)2=1n2rE(∑i=1naniXi)2≤K1n2r∑i=1nani2EXi2≤K2n2rEX2∑i=1nani2≤K3n2r-1,n≥1.
Consequently, (9) holds true.
3. Applications to the Convergence of the State Observers of Linear-Time-Invariant Systems
In this section, we give the applications of Corollary 6 and Theorem 7 to study the convergence of the state observers of linear-time-invariant systems.
For t≥0, consider an MISO (multi-input-single-output) linear-time-invariant system
(25)x˙(t)=Ax(t)+Bu(t),y(t)=Cx(t),
where A∈Rm0×m0, B∈Rm0×m1, and C∈R1×m0 are known system matrices, and for t≥0, u(t)∈Rm1 is the control input, x(t)∈Rm0 is the state, and y(t)∈R is the system output. The initial state x(0) is unknown. We are interested in estimation of x(t), from some limited observations on y(t).
In our setup, the output y(t) is only measured at a sequence of sampling time instants {ti} with measured values γ(ti), and noise di(26)γ(ti)=y(ti)-di.
We would like to estimate the state x(t) from information on u(t), {ti}, and {γ(ti)}. In practical systems, the irregular sampling sequences {γ(ti)} can be generated by different means such as randomized sampling, event-triggered sampling, and signal quantization.
It is obvious that state estimation will not be possible if the system is not observable. Also, in this paper, dk is assumed to be martingale difference. We give the following assumption.
Assumption 9.
The system (25) is observable; that is, the observability matrix
(27)Wo′=[C′,(CA)′,…,(CAm0-1)′]
has full rank.
For both t>t0 and t<t0, the solution to system (25) can be expressed as
(28)x(t)=eA(t-t0)x(t0)+∫t0teA(t-τ)Bu(τ)dτ.
Suppose that {ti,1≤i≤n} is a sequence of sampling times. For ti≤tn, we have
(29)γ(ti)+di=y(ti)=CeA(ti-tn)x(tn)+C∫tntieA(ti-τ)Bu(τ)dτ.
Since the second term is known, it will be denoted by v(ti,tn)=C∫tntieA(ti-τ)Bu(τ)dτ. This leads to the observations
(30)CeA(ti-tn)x(tn)=γ(ti)-v(ti,tn)+di,1≤i≤n.
Then, (30) can be written as
(32)Φnx(tn)=Γn-Vn+Dn.
Suppose that Φn is full rank, which will be established later. Then, a least-squares estimate of x(tn) is given by
(33)x^(tn)=(Φn′Φn)-1Φn′(Γn-Vn).
Here, G′ denotes the transpose of G. From (32) and (33), the estimation error for x(tn) at sampling time tn is
(34)e(tn)=x^(tn)-x(tn)=(Φn′Φn)-1Φn′Dn=(1nrΦn′Φn)-11nrΦn′Dn
for some 1/2<r<1. For convergence analysis, one must consider a typical entry in (1/nr)Φn′Dn. By the Cayley Hamilton theorem (see Ogata [18]), the matrix exponential can be expressed by a polynomial function of A of order at most m0-1,
(35)eAt=α1(t)I+⋯+αm0(t)Am0-1,
where the time functions αi(t) can be derived by the Lagrange-Hermite interpolation method (see Ogata [18]). This implies that
(36)CeA(ti-tn)=[α1(ti-tn),…,αm0(ti-tn)][CCA⋮CAm0-1]=φ′(ti-tn)Wo,
where φ′(ti-tn)=[α1(ti-tn),…,αm0(ti-tn)] and Wo is the observability matrix.
Denote
(37)Ψn=[φ′(t1-tn)⋮φ′(0)].
Then
(38)Φn=ΨnWo,
which implies that
(39)1nrΦn′Φn=Wo′1nrΨn′ΨnWo,1nrΦn′Dn=1nrWo′Ψn′Dn.
As a result, for any r>0, one has
(40)e(tn)=(1nrΦn′Φn)-11nrΦn′Dn=Wo-1(1nrΨn′Ψn)-11nrΨn′Dn.
Under Assumption 9, W0-1 exists. Convergence results will be established by the following two sufficient conditions: (1/nr)Ψn′Dn→0 and (1/nr)Ψn′Ψn≥βI, for some β>0. So we need the following persistent excitation (PE) condition, which was used by Wang et al. [19] and Thanh et al. [20].
Assumption 10.
For some 1/2<r<1,
(41)β=infn≥1σmin(1nrΨn′Ψn)>0,
where σmin(H) is the small eigenvalue of H for a suitable symmetric H.
We can investigate the convergence of double-indexed summations of random variables form
(42)1nr∑i=1nanidi
for some 1/2<r<1. Here, {ani} is a triangular array of real numbers and {di} is a sequence of martingale differences. It can be seen that (42) is a special case of (7) in Corollary 6. The jth component of (1/nr)Ψn′Dn takes the form
(43)1nr∑i=1nαj(ti-tn)di,
where {αj(ti-tn)} is a triangular array of real numbers. The convergence analysis of (43) for e(tn) is a special case of (42) or (7) in Corollary 6.
Recently, Wang et al. [19] investigated the convergence analysis of the state observers of linear-time-invariant systems under ρ*-mixing sampling. Thanh et al. [20] studied the convergence analysis of double-indexed and randomly weighted sums of ρ*-mixing sequence and gave its application to state observers. For more related works, one can refer to [18–23] and the references therein.
As an application of Corollary 6 to the observers and state estimation, we obtain the following theorem.
Theorem 11.
Let Assumptions 9 and 10 hold. Let 1/2<r<1 and {dn,ℱn,n≥1} be martingale differences stochastically dominated by a nonnegative random variable d with Ed2/r<∞. Suppose that for any q>2/(2r-1), one has E[supn≥1E(dn2∣ℱn-1)]q/2<∞ and
(44)∑i=1n|αj(ti-tn)|q=O(n),
where 1≤j≤m0. Then
(45)1nr||Ψn′Dn||⟶0,a.s.
Consequently,
(46)e(tn)⟶0,a.s.
As an application to Theorem 7, we get the following result.
Theorem 12.
Let 1/2<r<1 and Assumptions 9 and 10 hold. Assume that {dn,ℱn,n≥1} are martingales differences stochastically dominated by a nonnegative random variable d with Ed2<∞. For 1≤j≤m0, it is supposed that
(47)∑i=1nαj2(ti-tn)=O(n).
Then
(48)ζ=supn≥1n2r-1Ee′(tn)e(tn)<∞.
Remark 13.
If we assume that, for each 1≤i≤n, {φ(ti-tn)} is uniformly bounded, then we can find that condition (44) holds for any q. On the other hand, similar to Theorems 11 and 12, Wang et al. [19] also obtained the convergence of the state observers with probability one and in mean square under ρ*-mixing sampling (see Theorems 4 and 5 of Wang et al. [19]). So Theorems 11 and 12 generalize the results of Wang et al. [19] to the case of martingale differences.
Proof of Theorem 11.
It can be seen that
(49)1nrΨn′Dn=[1nr∑i=1nα1(ti-tn)di⋮1nr∑i=1nαm0(ti-tn)di].
To prove (45), it suffices to look at the jth component
(50)1nr∑i=1nαj(ti-tn)di
of
(51)1nrΨn′Dn.
For any q>2/(2r-1), by E[supn≥1E(dn2∣ℱn-1)]q/2<∞ and (44), we can obtain (45) from Corollary 6 with l=1/r, ani=αj(ti-tn) in (43), and Xn=dn.
On the other hand, by Assumption 9, W0-1 exists, and by (41) in Assumption 10, ((1/nr)Ψn′Ψn)-1 exists and
(52)σmax((1nrΨn′Ψn)-1)≤1β,
where σmax(·) is the largest eigenvalue. Together with
(53)e(tn)=Wo-1(1nrΨn′Ψn)-11nrΨn′Dn
and (45), it follows (46).
Proof of Theorem 12.
For 1≤j≤m0, by (47), (8) holds. Applying Theorem 7 with ani=αj(ti-tn), Xn=dn, and 1/2<r<1, we obtain that for a typical term
(54)1nr∑i=1nαj(ti-tn)di
in (49),
(55)n2r-1E(1nr∑i=1nαj(ti-tn)di)2≤K1,n≥1.
Together with (49), (53), and (55), we obtain that
(56)n2r-1Ee′(tn)e(tn)≤m0K2<∞,
where K2 is a positive constant. Lastly, by (56), (48) holds true.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This work is supported by the NNSF of China (11171001, 11201001, and 11326172), Natural Science Foundation of Anhui Province (1208085QA03 and 1408085QA02), Higher Education Talent Revitalization Project of Anhui Province (2013SQRL005ZD), Academic and Technology Leaders to Introduction Projects of Anhui University, and Doctoral Research Start-up Funds Projects of Anhui University.
HsuP. L.RobbinsH.Complete convergence and the law of large numbers19473322531MR001985210.1073/pnas.33.2.25ZBL0030.20101ErdösP.On a Theorem of Hsu and Robbins1949202286291MR003071410.1214/aoms/1177730037ZBL0033.29001BaumL. E.KatzM.Convergence rates in the law of large numbers19651201108123MR019852410.1090/S0002-9947-1965-0198524-1ZBL0142.14802YuK. F.Complete convergence of weighted sums of martingale differences1990323393472-s2.0-004224504110.1007/BF01045165MR1046337ZBL0698.60035GhosalS.ChandraT. K.Complete convergence of martingale arrays19981136216312-s2.0-0032373438MR163337810.1023/A:1022646429754ZBL0913.60029StoicaG.Baum-Katz-Nagaev type results for martingales20073362148914922-s2.0-3454808341710.1016/j.jmaa.2007.03.012MR2353031ZBL1130.60020StoicaG.A note on the rate of convergence in the strong law of large numbers for martingales201138129109132-s2.0-7995592289510.1016/j.jmaa.2011.04.008MR2802125ZBL1237.60025WangX. J.HuS. H.YangW. Z.WangX. H.Convergence rates in the strong law of large numbers for martingale difference sequences201220121357249310.1155/2012/572493ZBL1253.60045YangW. Z.HuS. H.WangX. J.Complete convergence for moving average process of martingale differences201220121612849210.1155/2012/128492ZBL1261.60038GutA.2005Springer Science+Business MediaChenP.HuT.-C.VolodinA.Limiting behaviour of moving average processes under φ-mixing assumption20097911051112-s2.0-5694908555310.1016/j.spl.2008.07.026SungS. H.Moment inequalities and complete moment convergence20092009142-s2.0-7044939765410.1155/2009/271265271265ZBL1180.60019SungS. H.Complete convergence for weighted sums of ρ*-mixing random variables20102010132-s2.0-7795256511310.1155/2010/630608630608ZBL1193.60045SungS. H.Convergence of moving average processes for dependent random variables20114013236623762-s2.0-7995614578810.1080/03610921003797761ZBL1219.60034SungS. H.VolodinA.A note on the rate of complete convergence for weighted sums of arrays of Banach space valued random elements20112922822912-s2.0-7995193128810.1080/07362994.2011.548670ZBL1217.60007HuT.-C.RosalskyA.VolodinA.A Complete Convergence Theorem for Row Sums from Arrays of Rowwise Independent Random Elements in Rademacher Type p Banach Spaces20123023433532-s2.0-8485820591910.1080/07362994.2012.649630MR2891458ZBL1237.60022HallP.HeydeC. C.1980New York, NY, USAAcademic PressOgataK.20024thEnglewood Cliffs, NJ, USAPrentice-HallWangL. Y.LiC.YinG. G.GuoL.XuC.-Z.State observability and observers of linear-time-invariant systems under irregular sampling and sensor limitations20115611263926542-s2.0-8045515048810.1109/TAC.2011.2122570ThanhL. V.YinG. G.WangL. Y.State observers with random sampling times and convergence analysis of double-indexed and randomly weighted sums of mixing processes20114911061242-s2.0-7995229151010.1137/10078671XZBL1219.93139HuangD. W.GuoL.Estimation of nonstationary ARMAX models based on the Hannan-Rissanen method199018417291756MR107443210.1214/aos/1176347875ZBL0721.62091WangL. Y.YinG.ZhangJ. F.ZhaoY. L.2010Boston, Mass, USABirkhäuserWuF.YinG.WangL. Y.Moment exponential stability of random delay systems with two-time-scale Markovian switching2012136247624902-s2.0-8485813334410.1016/j.nonrwa.2012.02.013MR2927201ZBL1263.60062