On the rate of convergence in the global central limit theorem for random sums of independent random variables

We present upper bounds of the integral \( {\int}_{-\infty}^{\infty }{\left|x\right|}^l\left|\mathbf{P}\left\{{Z}_N<x\right\}-\varPhi (x)\right|\mathrm{d}x \) for 0 ≤ l ≤ 1 + δ, where 0 < δ ≤ 1, Φ(x) is a standard normal distribution function, and Z _N = \( {S}_N/\sqrt{\mathbf{V}{S}_N} \) is the normalized random sum with variance V S _N > 0 (S _N = X ₁ + · · · + X _N ) of centered independent random variables X ₁ ,X ₂ , . . . . The number of summands N is a nonnegative integer-valued random variable independent of X ₁ ,X ₂ , . . . .     

Exact tail asymptotics of the supremum of strongly dependent Gaussian processes over a random interval

Let $ \mathcal{T} $ be a positive random variable independent of a real-valued stochastic process $ \left\{ {X(t),t\geqslant 0} \right\} $ . In this paper, we investigate the asymptotic behavior of $ \mathrm{P}\left( {{\sup_{{t\in \left[ {0,\mathcal{T}} \right]}}}X(t)>u} \right) $ as u→∞ assuming that X is a strongly dependent stationary Gaussian process and $ \mathcal{T} $ has a regularly varying survival function at infinity with index λ ∈ [0, 1). Under asymptotic restrictions on the correlation function of the process, we show that $ \mathrm{P}\left( {{\sup_{{t\in \left[ {0,\mathcal{T}} \right]}}}X(t)>u} \right)={c^{\lambda }}\mathrm{P}\left( {\mathcal{T}>m(u)} \right)\left( {1+o(1)} \right) $ with some positive finite constant c and function m(·) defined in terms of the local behavior of the correlation function and the standard Gaussian distribution.     

Recurrent statistical experiments with persistent linear regression

Statistical experiments with persistent linear regression are considered as normalized sums $ {S_N}(k)=\frac{1}{N}\sum\nolimits_{r=1}^N {{\delta_r}(k),\;\;k\geq 0} $ , of jointly independent and identically distributed, for every $ r\geq 1 $ , random variables $ {\delta_r}(k),\;\;1\leq r\leq N,\;\;k\geq 0 $ , which can take two values ±1. The persistent linear regression means that the recurrence relation $$ E\left[ {{S_N}\left( {k+1} \right)|{S_N}(k)} \right]=c-2a{S_N}(k),\;k\geq 0, $$ holds. We investigate the stable mode (equilibrium) (Theorem 2.1) and propose a stochastic approximation of statistical tests as a discrete-time autoregression process with normal perturbations (Theorem 3.1).     

A self-normalized law of the iterated logarithm for the geometrically weighted random series

Let {X, X_n; n ≥ 0} be a sequence of independent and identically distributed random variables with EX=0, and assume that EX~2I(|X| ≤ x) is slowly varying as x →∞, i.e., X is in the domain of attraction of the normal law. In this paper, a self-normalized law of the iterated logarithm for the geometrically weighted random series Σ~∞_(n=0)β~nX_n(0 β 1) is obtained, under some minimal conditions.     

Randomly weighted sums of dependent subexponential random variables*

We consider the randomly weighted sums $ \sum\nolimits_{k = 1}^n {{\theta_k}{X_k},n \geqslant 1} $ , where $ \left\{ {{X_k},1 \leqslant k \leqslant n} \right\} $ are n real-valued random variables with subexponential distributions, and $ \left\{ {{\theta_k},1 \leqslant k \leqslant n} \right\} $ are other n random variables independent of $ \left\{ {{X_k},1 \leqslant k \leqslant n} \right\} $ and satisfying $ a \leqslant \theta \leqslant b $ for some $ 0 < a \leqslant b < \infty $ and all $ 1 \leqslant k \leqslant n $ . For $ \left\{ {{X_k},1 \leqslant k \leqslant n} \right\} $ satisfying some dependent structures, we prove that $$ {\text{P}}\left( {\mathop {{\max }}\limits_{1 \leqslant m \leqslant n} \sum\limits_{k = 1}^m {{\theta_k}{X_k} > x} } \right)\sim {\text{P}}\left( {\sum\limits_{k = 1}^m {{\theta_k}{X_k} > x} } \right)\sim {\text{P}}\left( {\mathop {{\max }}\limits_{1 \leqslant k \leqslant n} {\theta_k}{X_k} > x} \right)\sim \sum\limits_{k = 1}^m {{\text{P}}\left( {{\theta_k}{X_k} > x} \right)} $$ as x??????.     

On distribution of the norm for normal random elements in the space of continuous functions

We consider distributions of norms for normal random elements X in separable Banach spaces, in particular, in the space C(S) of continuous functions on a compact space S. We prove that, under some nondegeneracy condition, the functions $ {{\mathcal{F}}_X}=\left\{ {\mathrm{P}\left( {\left\| {X-z} \right\|\leqslant r} \right):\;z\in C(S)} \right\},\;r\geqslant 0 $ , are uniformly Lipschitz and that every separable Banach space B can be ε-renormed so that the family $ {{\mathcal{F}}_X} $ becomes uniformly Lipschitz in the new norm for any B-valued nondegenerate normal random element X.     

A nonclassical LIL for geometrically weighted series in Banach space

Let {X, Xn ; n ≥ 0} be a sequence of independent and identically distributed random variables, taking values in a separable Banach space (B,||·||) with topological dual B* . Considering the geometrically weighted series ξ(β) =∑∞n=0βnXn for 0 β 1, and a sequence of positive constants {h(n), n ≥ 1}, which is monotonically approaching infinity and not asymptotically equivalent to log log n, a limit result for(1-β2)1/2||ξ(β)||/(2h(1/(1-β2)))1/2 is achieved.     

Asymptotic properties for the loglog laws under positive association

Let {X _n : n ?? 1} be a strictly stationary sequence of positively associated random variables with mean zero and finite variance. Set $S_n = \sum\limits_{k = 1}^n {X_k }$ , $Mn = \mathop {\max }\limits_{k \leqslant n} \left| {S_k } \right|$ , n ?? 1. Suppose that $0 < \sigma ^2 = EX_1^2 + 2\sum\limits_{k = 2}^\infty {EX_1 X_k < \infty }$ . In this paper, we prove that if E|X ₁|^2+?? < for some ?? ?? (0, 1], and $\sum\limits_{j = n + 1}^\infty {Cov\left( {X_1 ,X_j } \right) = O\left( {n^{ - \alpha } } \right)}$ for some ?? > 1, then for any b > ?1/2 $$\mathop {\lim }\limits_{\varepsilon \searrow 0} \varepsilon ^{2b + 1} \sum\limits_{n = 1}^\infty {\frac{{(\log \log n)^{b - 1/2} }} {{n^{3/2} \log n}}} E\left\{ {M_n - \sigma \varepsilon \sqrt {2n\log \log n} } \right\}_ + = \frac{{2^{ - 1/2 - b} E\left| N \right|^{2(b + 1)} }} {{(b + 1)(2b + 1)}}\sum\limits_{k = 0}^\infty {\frac{{( - 1)^k }} {{(2k + 1)^{2(b + 1)} }}}$$ and $$\mathop {\lim }\limits_{\varepsilon \nearrow \infty } \varepsilon ^{ - 2(b + 1)} \sum\limits_{n = 1}^\infty {\frac{{(\log \log n)^b }} {{n^{3/2} \log n}}E\left\{ {\sigma \varepsilon \sqrt {\frac{{\pi ^2 n}} {{8\log \log n}}} - M_n } \right\}} _ + = \frac{{\Gamma (b + 1/2)}} {{\sqrt 2 (b + 1)}}\sum\limits_{k = 0}^\infty {\frac{{( - 1)^k }} {{(2k + 1)^{2b + 2} }}} ,$$ where x ₊ = max{x, 0}, N is a standard normal random variable, and ??(·) is a Gamma function.     

Circular law theorem for random Markov matrices

Let (X _jk ) _jk≥1 be i.i.d. nonnegative random variables with bounded density, mean m, and finite positive variance σ ². Let M be the n?× n random Markov matrix with i.i.d. rows defined by ${M_{jk}=X_{jk}/(X_{j1}+\cdots+X_{jn})}$ . In particular, when X ₁₁ follows an exponential law, the random matrix M belongs to the Dirichlet Markov Ensemble of random stochastic matrices. Let λ₁, . . . , λ _n be the eigenvalues of ${\sqrt{n}M}$ i.e. the roots in ${\mathbb{C}}$ of its characteristic polynomial. Our main result states that with probability one, the counting probability measure ${\frac{1}{n}\delta_{\lambda_1}+\cdots+\frac{1}{n}\delta_{\lambda_n}}$ converges weakly as n→∞ to the uniform law on the disk ${\{z\in\mathbb{C}:|z|\leq m^{-1}\sigma\}}$ . The bounded density assumption is purely technical and comes from the way we control the operator norm of the resolvent.     

Generalization of the steinhaus problem on the convergence of series with random signs

For series of random variables $\sum\limits_{k = 1}^\infty {a_k x_k }$ ,a _K ∈R ¹, {X _K } _K=1 ^∞ being an Ising system, i.e., for each n ≥ 2 the joint distribution of {X _K } _K=1 ⁿ has the form $$P_n (t_1 ,...,t_n ) = ch^{ - (n - 1)} J \cdot exp(J\sum\limits_{k - 1}^{n - 1} {t_k t_{k + 1} )\prod\limits_{k = 1}^n {\frac{1}{2}\delta (t_{k^{ - 1} }^2 ),J > 0} }$$ one obtains a criterion for almost everywhere convergence: $\sum\limits_{k = 1}^\infty {a_k^2< \infty }$ . The relation between the asymptotic behavior of large deviations of the sum and the rate of decrease of the sequence {a_k} of the coefficients is investigated.     

Limiting Behavior of High Order Correlations for Simple Random Sampling

For ${N = 1, 2,\ldots,}$ let S _N be a simple random sample of size n = n _N from a population A _N of size N, where ${0 \leq n \leq N}$ . Then with f _N =  n/N, the sampling fraction, and 1 _A the inclusion indicator that ${A \in S_N}$ , for any ${H \subset A_N}$ of size ${k \geq 0}$ , the high order correlations $${\rm Corr}(k) = E \big(\mathop{\Pi}\limits_{A \in H} ({\bf 1}_A - f_N )\big)$$ depend only on k, and if the sampling fraction ${f_N \rightarrow f}$ as ${N \rightarrow \infty}$ , then $$N^{k/2}{\rm Corr}(k) \rightarrow [f (f - 1)]^{k/2}EZ^{k}, k \,{\rm even}$$ and $$N^{(k+1)/2}{\rm Corr}(k) \rightarrow [f (f - 1)]^{(k-1)/2}(2f - 1)\frac{1}{3}(k - 1)EZ^{k+1}, k \,{\rm odd}$$ where Z is a standard normal random variable. This proves a conjecture given in [2].     

The tail behaviour of a random sum of subexponential random variables and vectors

Let $\left\{ X,X_{i},i=1,2,...\right\} $ denote independent positive random variables having common distribution function (d.f.) F(x) and, independent of X, let ν denote an integer valued random variable. Using X ₀=0, the random sum Z=∑ _i=0 ^ν X _i has d.f. $G(x)=\sum_{n=0}^{\infty }\Pr\{\nu =n\}F^{n\ast }(x)$ where F ^n?(x) denotes the n-fold convolution of F with itself. If F is subexponential, Kesten’s bound states that for each ε>0 we can find a constant K such that the inequality $$ 1-F^{n\ast }(x)\leq K(1+\varepsilon )^{n}(1-F(x))\, , \qquad n\geq 1,x\geq 0 \, , $$ holds. When F is subexponential and E(1 +ε) ^ν <∞, it is a standard result in risk theory that G(x) satisfies $$ 1 - G{\left( x \right)} \sim E{\left( \nu \right)}{\left( {1 - F{\left( x \right)}} \right)},\,\,x \to \infty \,\,{\left( * \right)} $$ In this paper, we show that (*) holds under weaker assumptions on ν and under stronger conditions on F. Stam (Adv. Appl. Prob. 5:308–327, 1973) considered the case where $ \overline{F}(x)=1-F(x)$ is regularly varying with index –α. He proved that if α>1 and $E{\left( {\nu ^{{\alpha + \varepsilon }} } \right)} < \infty $ , then relation (*) holds. For 0<α<1, it is sufficient that Eν<∞. In this paper we consider the case where $\overline{F}(x)$ is an O-regularly varying subexponential function. If the lower Matuszewska index $\beta (\overline{F})<-1$ , then the condition ${\text{E}}{\left( {\nu ^{{{\left| {\beta {\left( {\overline{F} } \right)}} \right|} + 1 + \varepsilon }} } \right)} < \infty$ is sufficient for (*). If $\beta (\overline{F} )>-1$ , then again Eν<∞ is sufficient. The proofs of the results rely on deriving bounds for the ratio $\overline{F^{n\ast }}(x)/\overline{F} (x)$ . In the paper, we also consider (*) in the special case where X is a positive stable random variable or has a compound Poisson distribution derived from such a random variable and, in this case, we show that for n≥2, the ratio $\overline{F^{n\ast }}(x)/\overline{F}(x)\uparrow n$ as x↑∞. In Section 3 of the paper, we briefly discuss an extension of Kesten’s inequality. In the final section of the paper, we discuss a multivariate analogue of (*).     

A tauberian theorem for random walk

LetX ₁,X ₂,... be independent random variables, all with the same distribution symmetric about 0; $$S_n = \sum\limits_{i = 1}^n {X_i } $$ It is shown that if for some fixed intervalI, constant 1<a≦2 and slowly varying functionM one has $$\sum\limits_{k = 1}^n {P\{ S_k \in I\} \sim \frac{{n^{1 - 1/\alpha } }}{{M(n)}}} (n \to \infty )$$ then theX _i belong to the domain of attraction of a symmetric stable law.     

The norm of polynomials in large random and deterministic matrices

Let ${{\bf X}_N =(X_1^{(N)}, \ldots, X_p^{(N)})}$ be a family of N × N independent, normalized random matrices from the Gaussian Unitary Ensemble. We state sufficient conditions on matrices ${{\bf Y}_N =(Y_1^{(N)}, \ldots, Y_q^{(N)})}$ , possibly random but independent of X _N , for which the operator norm of ${P({\bf X}_N, {\bf Y}_N, {\bf Y}_N^*)}$ converges almost surely for all polynomials P. Limits are described by operator norms of objects from free probability theory. Taking advantage of the choice of the matrices Y _N and of the polynomials P, we get for a large class of matrices the ??no eigenvalues outside a neighborhood of the limiting spectrum?? phenomena. We give examples of diagonal matrices Y _N for which the convergence holds. Convergence of the operator norm is shown to hold for block matrices, even with rectangular Gaussian blocks, a situation including non-white Wishart matrices and some matrices encountered in MIMO systems.     

Strong approximation of empirical process with independent but non-identically distributed random variables

Let{Y_t,t=1,2,…} be independent random variables with continuous distribution functionsF_i(y).For any y,dencte s=F_t(y)=1/t sum from i=1 to t F_i(y).The empirical process is defind by t~(-1/2)R(s,t) whereR(s,t)=t(1/t sum from i=1 to t I_((?)_t(Y_i)≤s)-s)=sum from i=1 to t I_(?)-ts=sum from i=1 to t I_(?)-(?)_t(y)=sum from i=1 to t I_(Y_(?)≤y)-sum from i=1 to t F_i(y).The purpose of this paper is to investigate the asymptotic properties of the empirical processR(s,t).We shall prove that for some integer sequence {t_k},there is a (?)-process (?)(s,t) such that(?)|R(s,t_k)-(?)(s,t_k)|=O(t_k~(1/2)(log t_k)~(-1/4)(log log t_k)~(1/2))a.s.where (?)(s,t) is a two-parameter Gaussian process defined in §1.     

Generalized absolute Hausdorff summability of orthogonal series

For 1≦k≦2 and a sequence $\gamma :={\{\gamma(n)\}}_{n=1}^{\infty}$ that is quasi β-power monotone decreasing with ${\beta>1-\frac{1}{k}}$ , we prove the |A,γ| _k summability of an orthogonal series, where A is either a regular or Hausdorff matrix. For ${\beta>-\frac{3}{4}}$ , we give a necessary and sufficient condition for |A,γ| _k summability, where A is Hausdorff matrix. Our sufficient condition for ${\beta>-\frac{3}{4}}$ is weaker than that of Kantawala [1], ${\beta>-\frac{1}{k}}$ for |E,q,γ| _k summability; and of Leindler [4], β>?1 for |C,α,γ| _k , ${\alpha<\frac{1}{4}}$ . Also, our result generalizes the result of Spevakov [6] for |E,q,1|₁ summability.     

Two families of approximations for the gamma function

In this paper, we establish two families of approximations for the gamma function: $$ \begin{array}{lll} {\varGamma}(x+1)&=\sqrt{2\pi x}{\left({\frac{x+a}{{\mathrm{e}}}}\right)}^x {\left({\frac{x+a}{x-a}}\right)}^{-\frac{x}{2}+\frac{1}{4}} {\left({\frac{x+b}{x-b}}\right)}^{\sum\limits_{k=0}^m\frac{{\beta}_k}{x^{2k}}+O{{\left(\frac{1}{x^{2m+2}}\right)}}},\\ {\varGamma}(x+1)&=\sqrt{2\pi x}\cdot(x+a)^{\frac{x}{2}+\frac{1}{4}}(x-a)^{\frac{x}{2}-\frac{1}{4}} {\left({\frac{x-1}{x+1}}\right)}^{\frac{x^2}{2}}\\ &\quad\times {\left({\frac{x-c}{x+c}}\right)}^{\sum\limits_{k=0}^m\frac{{\gamma}_k}{x^{2k}}+O{\left({\frac{1}{x^{2m+2}}}\right)}}, \end{array}$$ where the constants ${\beta }_k$ and ${\gamma }_k$ can be determined by recurrences, and $a$ , $b$ , $c$ are parameters. Numerical comparison shows that our results are more accurate than Stieltjes, Luschny and Nemes’ formulae, which, to our knowledge, are better than other approximations in the literature.     

Closure of some heavy-tailed distribution classes under random convolution*

In this paper, we consider the closure property of a random convolution $ \sum\nolimits_{n = 0}^\infty {{p_n}{F^*}^n} $ , where F is a heavy-tailed distribution on [0, ??), and p _n (n?=?0, 1, . . . ) are the local probabilities of a nonnegative integer-valued random variable. We obtain conditions under which the fact that distribution F belongs to the dominatedly varying-tailed class, long-tailed class, or to the intersection of these classes implies that $ \sum\nolimits_{n = 0}^\infty {{p_n}{F^*}^n} $ is in the same class.     

Asymptotic distribution of singular values of powers of random matrices

Let x be a complex random variable such that \( {\mathbf{E}}x = 0,\,{\mathbf{E}}{\left| x \right|^2} = 1 \), and \( {\mathbf{E}}{\left| x \right|^4} < \infty \). Let \( {x_{ij}},i,j \in \left\{ {1,2, \ldots } \right\} \), be independent copies of x. Let \( {\mathbf{X}} = \left( {{N^{ - 1/2}}{x_{ij}}} \right) \), 1≤i,j≤N, be a random matrix. Writing X ^? for the adjoint matrix of X, consider the product X ^m X ^?m with some m ∈{1,2,...}. The matrix X ^m X ^?m is Hermitian positive semidefinite. Let λ₁,λ₂,...,λ _N be eigenvalues of X ^m X ^?m (or squared singular values of the matrix X ^m ). In this paper, we find the asymptotic distribution function \( {G^{(m)}}(x) = {\lim_{N \to \infty }}{\mathbf{E}}F_N^{(m)}(x) \) of the empirical distribution function \( F_N^{(m)}(x) = {N^{ - 1}}\sum\nolimits_{k = 1}^N {\mathbb{I}\left\{ {{\lambda_k} \leqslant x} \right\}} \), where \( \mathbb{I}\left\{ A \right\} \) stands for the indicator function of an event A. With m=1, our result turns to a well-known result of Marchenko and Pastur [V. Marchenko and L. Pastur, The eigenvalue distribution in some ensembles of random matrices, Math. USSR Sb., 1:457–483, 1967].