Estimates for the convergence rate in the central limit theorem for randomized separable statistics in a multinomial scheme

Assume that there are given random vectors ¯v₁, ..., ¯v_s, independent in their totality, where v_j=(v_1j, ... v_Nj) has the multinomial distribution M(nj, p1j, ..., P_Nj,. Assume further that _jm ^(N) (x₁, ..., x_s), j=1–k, k1 are random functions of s nonnegative integer arguments x₁, ..., x_s. One considers the multidimensional randomized separable statistic S_N=(S_N1, ..., S_Nk), where. One obtains estimates for the rate of convergence in the central limit theorem for S_N.Translated from Veroyatnostnye Raspredeleniya i Matematicheskaya Statistika, pp. 315–337, 1986.     

The strong approximation of extremal processes

Summary If X ₁, X ₂, ..., are i.i.d. random variables and Y _n =Max(X ₁, ..., X _n ); if for some sequences A _n , B_n, n=1, 2, ..., E _n (t)=A_nY_[nt]+B_n is such that E _n (1) weakly converges to a non degenerate limit distribution, then we prove that it is possible to construct a sequence of replicates of extremal processes E ⁽ⁿ⁾(t) on the same probability space, such that d(E _n (.), E ⁽ⁿ⁾(.))0 a.s., with the Levy metric. We give the rates of consistency of the approximations.     

Strong Approximation Theorems for Sums of Random Variables when Extreme Terms are Excluded

Let {X _n ;n≥1} be a sequence of i.i.d. random variables and let X ^(r) _n = X _j if |X _j | is the r-th maximum of |X ₁|, ..., |X _n |. Let S _n = X ₁+⋯+X _n and ^(r) S _n = S _n −(X ⁽¹⁾ _n +⋯+X ^(r) _n ). Sufficient and necessary conditions for ^(r) S _n approximating to sums of independent normal random variables are obtained. Via approximation results, the convergence rates of the strong law of large numbers for ^(r) S _n are studied. Received March 22, 1999, Revised November 6, 2000, Accepted March 16, 2001     

NA随机场对数律的收敛速度

设d是一个正整数, N ^d是d -维正整数格点.设{X_n , n∈N ^d} 是一同分布的负相伴随机场, 记S_n =∑_{k≤ n} X_k, S_n^(k)=S_n-X_k, 如果r >2, EX₁ = 0 和σ²= Var(X₁}, 则存在一个正数M:=100√(r-2)(1+σ²)使得下列条件等价 (I)E |X₁|^r (log|X₁|)^d-1-r/2 <∞; (II)∑_{n∈ N^d} |n|^r/2-2P(max_{1≤ k≤ n} |S_n^(k)|≥ (2^d+1 )ε√|n| log |n |) <∞,∨ε > M; (III)∑_{n∈N ^d} |n|^r/2-2P(max_{1≤ k≤n} |S_k |≥ε√| n} log| n |) <∞,∨ε > M. (III)\ \ $\sum\limits_{{{\bf n}}\in {{\cal N}}^{d}} |n|^{r/2-2} P(\max\limits_{{\bf 1}\leq{\bf k}\leq{\bf n}}|S_{{\bf k}}|\geq \varepsilon \sqrt{|{\bf n}|\log |{\bf n}|})<\infty$, $\forall\varepsilon>M$.     

Random symmetric polynomials

Let {Q⁽ⁿ⁾(x₁,...,x_n)} be a sequence of symmetric polynomials having a fixed degree equal to k. Let {X_n1,...,X_nn}, n k, be some sequence of series of random variables (r.v.). We form the sequence of r.v. Y_n=Q⁽ⁿ⁾(X_n1, ... X_nn), n k One obtains limit theorems for the sequence Y_n, under very general assumptions.Translated from Veroyatnostnye Raspredeleniya i Matematicheskaya Statistika, pp. 170–188, 1986.     

The cyclic structure of random permutations

Letα _r denote the number of cycles of length r in a random permutation, taking its values with equal probability from among the set S_n of all permutations of length n. In this paper we study the limiting behavior of linear combinations of random permutationsα ₁, ...,α _r having the form $$\zeta _{n, r} = c_{r1^{a_1 } } + ... + c_{rr} a_r $$ in the case when n, r→∞. We shall show that the class of limit distributions forξ _n,r as n, r→∞ and r In r/h→0 coincides with the class of unbounded divisible distributions. For the random variables η_n, r=α ₁+2α ₂+... rα _r, equal to the number of elements in the permutation contained in cycles of length not exceeding r, we find' limit distributions of the form r In r/n→0 and r=γ ⁿ, 0<γ<1.     

A generalization of a theorem of Amemiya and Ando on the convergence of random products of contractions in Hilbert space

Let {T₁, ..., T_N} be a finite set of linear contraction mappings of a Hilbert space H into itself, and let r be a mapping from the natural numbersN to {1, ..., N} which assumes each value infinitely often. One can form S_n=T_r(n)...T_r(1) which could be described as a random product of the T_i's. If the contractions have the condition (W): Tx<x whenever Txx, then S_n converges weakly to the projection Q onto the subspace . This theorem is due to Amemiya and Ando. We demonstrate a basic property of the algebraic semigroupS=S(T₁, ..., T_N) generated by N contractions, each having (W). We prove that if the semigroup of an infinite set of contractions is equipped with this property, and the maps satisfy a minor condition parallel to (W) on each of N maps, then random products still converge weakly. Our proof is different from Amemiya and Ando's. We illustrate our method with a new proof of the fact that if a contraction T is completely non-unitary, then Tⁿ0 weakly.     

Elementary symmetric polynomials of increasing order

Summary The asymptotic behaviour of elementary symmetric polynomials S _n ^(k) of order k, based on n independent and identically distributed random variables X ₁,..., X _n,is investigated for the case that both k and n get large. If , then the distribution function of a suitably normalised S _n ^(k) is shown to converge to a standard normal limit. The speed of this convergence to normality is of order , provided and certain natural moment assumptions are imposed. This order bound is sharp, and cannot be inferred from one of the existing Berry-Esseen bounds for U-statistics. If k at the rate n ^1/2 then a non-normal weak limit appears, provided the X _i's are positive and S _n ^(k) is standardised appropriately. On the other hand, if k at a rate faster than n ^1/2 then it is shown that for positive X _j'sthere exists no linear norming which causes S _n ^(k) to converge weakly to a nondegenerate weak limit.     

General theorems on rates of convergence in distribution of random variables I. General limit theorems

Let (X_n)_n?$\text{N}$ be a sequence of real, independent, not necessarily identically distributed random variables (r.v.) with distribution functions F_Xn, and S_n = Σ_i=1ⁿX_i. The authors present limit theorems together with convergence rates for the normalized sums ?(n)S_n, where ?: $\text{N}$ → $\text{R}$⁺, ?(n) → 0, n → ∞, towards appropriate limiting r.v. X, the convergence being taken in the weak (star) sense. Thus higher order estimates are given for the expression ∝_$\text{R}$f(x) d[F_?(n)Sn(x) ? F_X(x)] which depend upon the normalizing function ?, decomposability properties of X and smoothness properties of the function f under consideration. The general theorems of this unified approach subsume O- and o-higher order error estimates based upon assumptions on associated moments. These results are also extended to multi-dimensional random vectors.     

A class of limit distributions for maximum cumulative sum

Let X₁, X₂, ... be a sequence of independent identically distributed random variables with zero mathematical expectation and finite variances. S_o=0 and S_n=∑ _i=1 ⁿ X_i. It is proved that is the limit distribution function of the normalized random variable a(k, n)} for some sequence of centering constants a (k,n).     

Almost sure central limit theorems for random functions

Let {Xn,-∞< n <∞} be a sequence of independent identically distributed random variables with EX1 = 0, EX12 = 1 and let Sn =∑k=1∞Xk, and Tn = Tn(X1,…,Xn) be a random function such that Tn = ASn Rn, where supn E|Rn| <∞and Rn = o(n~(1/2)) a.s., or Rn = O(n1/2-2γ) a.s., 0 <γ< 1/8. In this paper, we prove the almost sure central limit theorem (ASCLT) and the function-typed almost sure central limit theorem (FASCLT) for the random function Tn. As a consequence, it can be shown that ASCLT and FASCLT also hold for U-statistics, Von-Mises statistics, linear processes, moving average processes, error variance estimates in linear models, power sums, product-limit estimators of a continuous distribution, product-limit estimators of a quantile function, etc.     

A characteristic property of spheres

Summary We prove: Let S be a closed n-dimensional surface in an(n+1)-space of constant curvature (n ≥ 2); k₁ ≥ ... ≥ k_n denote its principle curvatures. Let φ(ξ₁, ..., ξ_n) be such that . Then if φ(k₁, ..., k_n)=const on S and S is subject to some additional general conditions (those(II ₀) or(II) n^o 1), S is a sphere. To Enrico Bompiani on his scientific Jubilee     

Reconstituting the distribution of the components of a random Vector from distributions of certain statistics

Let (X₁, ..., X_n) be a random vector with independent components. It is proven in this paper that, under certain restrictions, the distributions of the pairS ₁=sup (a ₁X₁, ..., a_nX_n) andS ₂=sup (b₁X₁,...,b_nX_n) univocally define the distribution function of the components X_j.Translated from Matematicheskie Zametki, Vol. 13, No. 6, pp. 889–892, June, 1973.     

On Strong Versions of the Central Limit Theorem

Let S_n be the sum of n i.i.d.r.v. and let 1_(-∞,x)(·) be the indicator function of the interval (-∞, x). Then the sequence 1_{(-∞, x)}(S_n/√n) does not converge for any x. Likewise the arithmetic means of this sequence converge only with probability zero. But the logarithmic means converge with probability one to the standard normal distribution Ø(x). Then for any bounded and a.e. continuous function a(y) the logarithmic means of a(S_n/√n) converge a.s. to a = ∫a(y)dØ(y). The arithmetic means of a(S_nk/√n) converge to the same limit a for all subsequences n_k = [c^k], c > 1.     

Asymptotic behavior of self-normalized trimmed sums: Nonnormal limits II

Let {X _j} be independent, identically distributed random variables which are symmetric about the origin and have a continuous nondegenerate distributionF. Let {X _n(1),...,X _n(n)} denote the arrangement of {X ₁,...,X _n} in decreasing order of magnitude, so that with probability one, |X _n(1)|>|X _n(2)|>...> |X _n(n)|. For initegersr _n such thatr _n/n0, define the self-normalized trimmed sumT _n= _i=rn ⁿ X _n(i)/{ _i=rn ⁿ X _n ² (i)}^1/2. Hahn and Weiner⁽⁶⁾ showed that under a probabilistically meaningful analytic condition generalizing the asymptotic normality criterion forT _n, various nonnormal limit laws forT _n arise which are represented by means of infinite random series. The analytic condition is now extended and the previous approach is refined to obtain limits which are mixtures of a normal, a Rademacher, and a law represented by a more general random series. Each such limit law actually arises as can be seen from the construction of a single distribution whose correspondingL(T _n ) generates all of the law along different subsequences, at least if {r _n} grows sufficiency fast. Another example clarifies the limitations of the basic approach.     

Large-deviation probability and the local dimension of sets

Let X ₁ , X ₂ , ..., X_n be n independent identically distributed real random variables and S_n = Σ _n=1 ⁿ X_i. We obtain precise asymptotics forP (S_n ∈ nA) for rather arbitrary Borel sets A₁ in terms of the density of the dominating points in A. Our result extends classical theorems in the field of large deviations for independent samples. We also obtain asymptotics forP (S_n ∈ γ_nA), with γ_n/n → ∞. Proceedings of the Seminar on Stability Problems for Stochastic Models, Vologda, Russia, 1998, Part I.     

Points of increase for random walks

Say that a sequenceS ₀, ..., S_n has a (global) point of increase atk ifS _k is maximal amongS ₀, ..., S_k and minimal amongS _k, ..., S_n. We give an elementary proof that ann-step symmetric random walk on the line has a (global) point of increase with probability comparable to 1/logn. (No moment assumptions are needed.) This implies the classical fact, due to Dvoretzky, Erdős and Kakutani (1961), that Brownian motion has no points of increase. Research partially supported by NSF grant # DMS-9404391.     

On P.I. ring and standard identities

An example is given of a ringR (with 1) satisfying the standard identityS ₆[x ₁, ...,x ₆] butM ₂(R), the 2 × 2 matrix ring overR, does not satisfyS ₁₂[x ₁, ...,x ₁₂]. This is in contrast to the caseR=M _n (F),F a field, where by the Amitsur-Levitzki theoremR satisfiesS _2n [x ₁, ...,x _2n] andM ₂(R) satisfiesS _4n [x ₁, ...,x _n]. Part of this work was done while the author enjoyed the hospitality of the University of California at San Diego, the University of Texas at Austin and the University of Washington at Seattle.     

When isf(x1, x2, ..., xn) =u1(x1) +u2(x2) + ... +un(xn)?

We discuss subsetsS of ℝⁿ such that every real valued functionf onS is of the formf(x₁, x₂, ..., x_n) =u ₁(x₁) +u ₂(x₂) +...+u _n(x_n), and the related concepts and situations in analysis.     

On extremal sort sequences

A sort sequence S_n is a sequence of all unordered pairs of indices inI _n = {1, 2, ..., n}. With a sort sequenceS_n = (s₁ ,s₂ ,...,s_{( 2ⁿ )} )S_n = (s_1 ,s_2 ,...,s_{\left( {_2^n } \right)} ), one can associate a predictive sorting algorithm A(S_n). An execution of the algorithm performs pairwise comparisons of elements in the input setX in the order defined by the sort sequence S_n except that the comparisons whose outcomes can be inferred from the results of the preceding comparisons are not performed. A sort sequence is said to be extremal if it maximizes a given objective function. First we consider the extremal sort sequences with respect to the objective function ω(S_n) — the expected number of active predictions inS _n. We study ω-extremal sort sequences in terms of their prediction vectors. Then we consider the objective function Ω(S_n) — the minimum number of active predictions in S_n over all input orderings.