Limit Theory for Moderate Deviations from a Unit Root Under Innovations with a Possibly Infinite Variance

An asymptotic theory was given by Phillips and Magdalinos (J Econom 136(1):115–130, 2007) for autoregressive time series Y _t ?=?ρY _t?1?+?u _t , t?=?1,...,n, with ρ?=?ρ _n ?=?1?+?c/k _n , under (2?+?δ)-order moment condition for the innovations u _t , where δ?>?0 when c?<?0 and δ?=?0 when c?>?0, {u _t } is a sequence of independent and identically distributed random variables, and (k _n ) _n?∈?? is a deterministic sequence increasing to infinity at a rate slower than n. In the present paper, we established similar results when the truncated second moment of the innovations $l(x)=\textsf{E} [u_1^2I\{|u_1|\le x\}]$ is a slowly varying function at ∞, which may tend to infinity as x?→?∞. More interestingly, we proposed a new pivotal for the coefficient ρ in case c?<?0, and formally proved that it has an asymptotically standard normal distribution and is nuisance parameter free. Our numerical simulation results show that the distribution of this pivotal approximates the standard normal distribution well under normal innovations.     

On shortest crucial words avoiding abelian powers

Let k≥2 be an integer. An abeliankth power is a word of the form X₁X₂?X_k where X_i is a permutation of X₁ for 2≤i≤k. A word W is said to be crucial with respect to abelian kth powers if W avoids abelian kth powers, but Wx ends with an abelian kth power for any letter x occurring in W.Evdokimov and Kitaev (2004) [2] have shown that the shortest length of a crucial word on n letters avoiding abelian squares is 4n−7 for n≥3. Furthermore, Glen et al. (2009) [3] proved that this length for abelian cubes is 9n−13 for n≥5. They have also conjectured that for any k≥4 and sufficiently large n, the shortest length of a crucial word on n letters avoiding abelian kth powers, denoted by ?_k(n), is k²n−(k²+k+1). This is currently the best known upper bound for ?_k(n), and the best known lower bound, provided in Glen et al., is 3kn−(4k+1) for n≥5 and k≥4. In this note, we improve this lower bound by proving that for n≥2k−1, ?_k(n)≥k²n−(2k³−3k²+k+1); thus showing that the aforementioned conjecture is true asymptotically (up to a constant term) for growing n.     

Rainbow paths

A k-rainbow path in a graph with colored edges is a path of length k where each edge has a different color. In this note, we settle the problem of obtaining a constructive k-coloring of the edges of K_n in which one may find, between any pair of vertices, a large number of internally disjoint k-rainbow paths. In fact, our construction obtains the largest possible number of paths. This problem was considered in a less general setting by Chartrand et al. (2007) [6].     

A semi-Markov process with an inverse Gaussian distribution as sojourn time

Let X_n denote the state of a device after n repairs. We assume that the time between two repairs is the time τ taken by a Wiener process {W(t), t ? 0}, starting from w₀ and with drift μ < 0, to reach c ∈ [0, w₀). After the nth repair, the process takes on either the value X_n?1 + 1 or X_n?1 + 2. The probability that X_n = X_n?1 + j, for j = 1, 2, depends on whether τ ? t₀ (a fixed constant) or τ > t₀. The device is considered to be worn out when X_n ? k, where k ∈ {1, 2, …}. This model is based on the ones proposed by Rishel (1991) [1] and Tseng and Peng (2007) [2]. We obtain an explicit expression for the mean lifetime of the device. Numerical methods are used to illustrate the analytical findings.     

Matroids with few non-common bases

In [On Mills's conjecture on matroids with many common bases, Discrete Math. 240 (2001) 271-276], Lemos proved a conjecture of Mills [On matroids with many common bases, Discrete Math. 203 (1999) 195-205]: for two (k+1)-connected matroids whose symmetric difference between their collections of bases has size at most k, there is a matroid that is obtained from one of these matroids by relaxing n₁ circuit-hyperplanes and from the other by relaxing n₂ circuit-hyperplanes, where n₁ and n₂ are non-negative integers such that n₁+n₂≤k. In [Matroids with many common bases, Discrete Math. 270 (2003) 193-205], Lemos proved a similar result, where the hypothesis of the matroids being k-connected is replaced by the weaker hypothesis of being vertically k-connected. In this paper, we extend these results.     

Splitting multidimensional necklaces

The well-known “splitting necklace theorem” of Alon [N. Alon, Splitting necklaces, Adv. Math. 63 (1987) 247-253] says that each necklace with k⋅ai beads of color i=1,…,n, can be fairly divided between k thieves by at most n(k−1) cuts. Alon deduced this result from the fact that such a division is possible also in the case of a continuous necklace [0,1] where beads of given color are interpreted as measurable sets Ai⊂[0,1] (or more generally as continuous measures μi). We demonstrate that Alon's result is a special case of a multidimensional consensus division theorem about n continuous probability measures μ₁,…,μn on a d-cube d[0,1]. The dissection is performed by m₁+?+md=n(k−1) hyperplanes parallel to the sides of d[0,1] dividing the cube into m₁⋅?⋅md elementary cuboids (parallelepipeds) where the integers mi are prescribed in advance.     

Thue-Morse at multiples of an integer

Let t=(tn)_n?0 be the classical Thue-Morse sequence defined by , where s₂ is the sum of the bits in the binary representation of n. It is well known that for any integer k?1 the frequency of the letter “1” in the subsequence t₀,tk,t2k,… is asymptotically 1/2. Here we prove that for any k there is an n?k+4 such that tkn=1. Moreover, we show that n can be chosen to have Hamming weight ?3. This is best in a twofold sense. First, there are infinitely many k such that tkn=1 implies that n has Hamming weight ?3. Second, we characterize all k where the minimal n equals k, k+1, k+2, k+3, or k+4. Finally, we present some results and conjectures for the generalized problem, where s₂ is replaced by sb for an arbitrary base b?2.     

Type-B generalized triangulations and determinantal ideals

For n≥3, let Ω_n be the set of line segments between the vertices of a convex n-gon. For j≥2, a j-crossing is a set of j line segments pairwise intersecting in the relative interior of the n-gon. For k≥1, let Δ_n,k be the simplicial complex of (type-A) generalized triangulations, i.e. the simplicial complex of subsets of Ω_n not containing any (k+1)-crossing.The complex Δ_n,k has been the central object of many papers. Here we continue this work by considering the complex of type-B generalized triangulations. For this we identify line segments in Ω_2n which can be transformed into each other by a 180^°-rotation of the 2n-gon. Let F_n be the set Ω_2n after identification, then the complex D_n,k of type-B generalized triangulations is the simplicial complex of subsets of F_n not containing any (k+1)-crossing in the above sense. For k=1, we have that D_n,1 is the simplicial complex of type-B triangulations of the 2n-gon as defined in [R. Simion, A type-B associahedron, Adv. Appl. Math. 30 (2003) 2-25] and decomposes into a join of an (n−1)-simplex and the boundary of the n-dimensional cyclohedron. We demonstrate that D_n,k is a pure, k(n−k)−1+kn dimensional complex that decomposes into a kn−1-simplex and a k(n−k)−1 dimensional homology-sphere. For k=n−2 we show that this homology-sphere is in fact the boundary of a cyclic polytope. We provide a lower and an upper bound for the number of maximal faces of D_n,k.On the algebraical side we give a term order on the monomials in the variables X_ij,1≤i,j≤n, such that the corresponding initial ideal of the determinantal ideal generated by the (k+1) times (k+1) minors of the generic n×n matrix contains the Stanley-Reisner ideal of D_n,k. We show that the minors form a Gröbner-Basis whenever k∈{1,n−2,n−1} thereby proving the equality of both ideals and the unimodality of the h-vector of the determinantal ideal in these cases. We conjecture this result to be true for all values of k<n.     

A combinatorial result related to the consistency of New Foundations

We prove a combinatorial result for models of the 4-fragment of the Simple Theory of Types (TST), TST₄. The result says that if A=〈A₀,A₁,A₂,A₃〉 is a standard transitive and rich model of TST₄, then A satisfies the 〈0,0,n〉-property, for all n≥2. This property has arisen in the context of the consistency problem of the theory New Foundations (NF). The result is a weak form of the combinatorial condition (existence of ω-extendible coherent triples) that was shown in Tzouvaras (2007) [5] to be equivalent to the consistency of NF. Such weak versions were introduced in Tzouvaras (2009) [6] in order to relax the intractability of the original condition. The result strengthens one of the main theorems of Tzouvaras (2007) [5, Theorem 3.6] which is just equivalent to the 〈0,0,2〉-property.     

On the modular sumset partition problem

A sequence m₁≥m₂≥?≥m_k of k positive integers isn-realizable if there is a partition X₁,X₂,…,X_k of the integer interval [1,n] such that the sum of the elements in X_i is m_i for each i=1,2,…,k. We consider the modular version of the problem and, by using the polynomial method by Alon (1999) [2], we prove that all sequences in Z/pZ of length k≤(p−1)/2 are realizable for any prime p≥3. The bound on k is best possible. An extension of this result is applied to give two results of p-realizable sequences in the integers. The first one is an extension, for n a prime, of the best known sufficient condition for n-realizability. The second one shows that, for n≥(4k)³, an n-feasible sequence of length k isn-realizable if and only if it does not contain forbidden subsequences of elements smaller than n, a natural obstruction forn-realizability.     

Functional central limit theorems for sums of nearly nonstationary processes*

We study some Holderian functional central limit theorems for the polygonal partial-sum processes built on a first-order autoregressive process y _n,k ?=??? _n y _n,k?1?+??? _k with ? _n converging to 1 and i.i.d. centered square-integrable innovations. In the case where ? _n ?=?e ^??/n with a negative constant ??, we prove that the limiting process is an integrated Ornstein?CUhlenbeck one. In the case where ? _n ?=?1? ?? _n /n, with ?? _n tending to infinity slower than n, the convergence to Brownian motion is established in Holder space in terms of the rate of ?? _n and the integrability of the ?? _k s.     

Matrix characterization of linear codes with arbitrary Hamming weight hierarchy

The support of an [n, k] linear code C over a finite field F_q is the set of all coordinate positions such that at least one codeword has a nonzero entry in each of these coordinate position. The rth generalized Hamming weight d_r(C), 1 ⩽ r ⩽ k, of C is defined as the minimum of the cardinalities of the supports of all [n, r] subcodes of C. The sequence (d₁(C), d₂(C), … , d_k(C)) is called the Hamming weight hierarchy (HWH) of C. The HWH, d_r(C) = n − k + r;  r = 1, 2 , …, k, characterizes maximum distance separable (MDS) codes. Therefore the matrix characterization of MDS codes is also the characterization of codes with the HWH d_r(C) = n − k + r; r = 1, 2, … , k. A linear code C with systematic check matrix [I∣P], where I is the (n − k) × (n − k) identity matrix and P is a (n − k) × k matrix, is MDS iff every square submatrix of P is nonsingular. In this paper we extend this characterization to linear codes with arbitrary HWH. Using this result, we characterize Near-MDS codes, Near-Near-MDS (N²-MDS) codes and A^μ-MDS codes. The MDS-rank of C is the smallest integer η such that d_η+1 = n − k + η + 1 and the defect vector of C with MDS-rank η is defined as the ordered set {μ₁(C), μ₂(C), μ₃(C), … , μ_η(C), μ_η+1(C)}, where μ_i(C) = n − k + i − d_i(C). We call C a dually defective code if the defect vector of the code and its dual are the same. We also discuss matrix characterization of dually defective codes. Further, the codes meeting the generalized Greismer bound are characterized in terms of their generator matrix. The HWH of dually defective codes meeting the generalized Greismer bound are also reported.     

On the class of limits of lacunary trigonometric series

Let (n _k ) _k≧1 be a lacunary sequence of positive integers, i.e. a sequence satisfying n _k+1/n _k > q > 1, k ≧ 1, and let f be a “nice” 1-periodic function with ∝ ₀ ¹ f(x) dx = 0. Then the probabilistic behavior of the system (f(n _k x)) _k≧1 is very similar to the behavior of sequences of i.i.d. random variables. For example, Erd?s and Gál proved in 1955 the following law of the iterated logarithm (LIL) for f(x) = cos 2πx and lacunary $ (n_k )_{k \geqq 1} $ : (1) $$ \mathop {\lim \sup }\limits_{N \to \infty } (2N\log \log N)^{1/2} \sum\limits_{k = 1}^N {f(n_k x)} = \left\| f \right\|_2 $$ for almost all x ∈ (0, 1), where ‖f‖₂ = (∝ ₀ ¹ f(x)² dx)^1/2 is the standard deviation of the random variables f(n _k x). If (n _k ) _k≧1 has certain number-theoretic properties (e.g. n _k+1/n _k → ∞), a similar LIL holds for a large class of functions f, and the constant on the right-hand side is always ‖f‖₂. For general lacunary (n _k ) _k≧1 this is not necessarily true: Erd?s and Fortet constructed an example of a trigonometric polynomial f and a lacunary sequence (n _k ) _k≧1, such that the lim sup in the LIL (1) is not equal to ‖f‖₂ and not even a constant a.e. In this paper we show that the class of possible functions on the right-hand side of (1) can be very large: we give an example of a trigonometric polynomial f such that for any function g(x) with sufficiently small Fourier coefficients there exists a lacunary sequence (n _k ) _k≧1 such that (1) holds with √‖f‖ ₂ ² + g(x) instead of ‖f‖₂ on the right-hand side.     

Some dimensional results for homogeneous Moran sets

Let ?(¦n _k ¦k?1,¦c _k ¦k?1) be the collection of homogeneous Moran sets determined by ¦n _k ¦k?1 and ¦c _k ¦k?1, where ¦n _k ¦k?1 is a sequence of positive integers and ¦c _k ¦k?1 a sequence of positive numbers. Then the maximal and minimal values of Hausdorff dimensions for elements in ? are determined. The result is proved that for any values between the maximal and minimal values, there exists an element in ?(¦n _k ¦k?1,¦c _k¦k?1) such that its Hausdorff dimension is equal tos. The same results hold for packing dimension. In the meantime, some other properties of homogeneous Moran sets are discussed.     

An exact algorithm for the knapsack sharing problem with common items

We are concerned with a variation of the knapsack problem as well as of the knapsack sharing problem, where we are given a set of n items and a knapsack of a fixed capacity. As usual, each item is associated with its profit and weight, and the problem is to determine the subset of items to be packed into the knapsack. However, in the problem there are s players and the items are divided into s + 1 disjoint groups, N_k (k = 0, 1, … , s). The player k is concerned only with the items in N₀ ∪ N_k, where N₀ is the set of ‘common’ items, while N_k represents the set of his own items. The problem is to maximize the minimum of the profits of all the players. An algorithm is developed to solve this problem to optimality, and through a series of computational experiments, we evaluate the performance of the developed algorithm.     

A Bijective Approach to the Area of Generalized Motzkin Paths

For fixed positive integer k, let E_n denote the set of lattice paths using the steps (1, 1), (1, − 1), and (k, 0) and running from (0, 0) to (n, 0) while remaining strictly above the x-axis elsewhere. We first prove bijectively that the total area of the regions bounded by the paths of E_n and the x-axis satisfies a four-term recurrence depending only on k. We then give both a bijective and a generating function argument proving that the total area under the paths of E_n equals the total number of lattice points on the x-axis hit by the unrestricted paths running from (0, 0) to (n − 2, 0) and using the same step set as above.     

Some exact distributions of the number of one-sided deviations and the time of the last such deviation in the simple random walk

For every positive integer n, let S_n be the n-th partial sum of a sequence of independent and identically distributed random variables, each assuming the values +1 and −1 with respective probabilities p (0<p<1)) and q (= 1 −p) and having mean μ = p − q. For a fixed positive real number λ, let N⁺[N1] be the total number of values of n for which S_n > (μ + λ)n [S_n⩾(μ + λ)n] and let L⁺[L1] be the supremum of the values of n for which S_n > (μ + λ)n [S_n⩾(μ + λ)n], where sup Oslash; = 0. Explicit expressions for the exact distributions of N⁺, N1, L⁺ and L1 are given when μ + λ = ±k/(k + 2) for any nonnegative integer k.     

Distribution of roots of quasianalytic functions

For functions of certain quasianalytic classes C{m_n} on (?∞, ∞) we determine a function ξ (x), depending on {m_n}, which is such that a sequence {x_k} is a sequence of the roots off(x) ε C{m_n} if and only if for somea $$\int_a^\infty {\tfrac{{dn(x)}}{{\xi (x - a}}< \infty ,} $$ where n(x) is a distribution function of the sequence {x_k}.     

Limit distributions in metric discrepancy theory

Let (n _k ) _{k ≥ 1} be a lacunary sequence of integers, satisfying certain number-theoretic conditions. We determine the limit distribution of ${\sqrt{N} D_N (n_{k} x)}$ as ${N \to \infty}$ , where D _N (n _k x) denotes the discrepancy of the sequence (n _k x) _{k ≥ 1} mod 1.