On tail parameter estimation in certain point process models

Let μ(ds, dx) denote Poisson random measure with intensity dsG(dx) on (0, ∞) × (0, ∞), for a measure G(dx) with tails varying regularly at ∞. We deal with estimation of index of regular variation α and weight parameter ξ if the point process is observed in certain windows K_n = [0, T_n] × [Y_n, ∞), where Y_n → ∞ as n → ∞. In particular, we look at asymptotic behaviour of the Hill estimator for α. In certain submodels, better estimators are available; they converge at higher speed and have a strong optimality property. This is deduced from the parametric case G(dx) = ξαx^−α−1 dx via a neighbourhood argument in terms of Hellinger distances.     

Estimation in dynamic regression with an integrated process

This paper deals with a dynamic regression model y_t = αy_t−1 + βz_t + u_t, where z_t is an integrated process of order one abbreviated as z_t ∼ I(1). Generally speaking, nonstandard asymptotic theory is required to investigate asymptotic properties of statistics related to an integrated process and the asymptotic results are very different from standard ones. There are two distinctive properties in nonstandard asymptotics: the so-called ‘super-consistency’ or T-consistency (where T is a sample size) and the weak convergence to a functional of the Wiener process. In spite of z_t being involved in our model, however, it is shown that our asymptotic results are the same as in the standard asymptotics in classical dynamic regression models, or if the disturbance u_t is serially correlated the OLS estimators of α and β have √T-inconsistency. This is due to the cointegration between y_t−1 and z_t. Although this point was clarified by Park and Phillips (1989) in a general context, we examine this explicitly through our specific model and connect the standard asymptotic theory with the nonstandard one in our case. Furthermore we investigate the limiting properties of other statistics such as t-ratio, the Durbin-Watson test and h-test. We also propose a consistent estimator of α and β by making use of Durbin's 2-step method. Finally, we carry out simulation studies which support our theoretical results.     

Constructions for point-regular linear spaces

We introduce the concept of simple difference family over a group G and relative to a partial spread in G. Such a family generates a point-regular linear space, i.e. a linear space with an automorphism group acting regularly on the point-set. In particular, we prove that any abelian linear space is generated by such a family. Using this new notion of difference family, we give a number of recursive constructions for point-regular linear spaces.     

Arrays of strength s on two symbols

This paper gives necessary and sufficient conditions on σ, s, t and on μ, s, t for an array with s+t rows to have strength s and weight σ, or to be balanced and have strength s and weight μ. If a balanced array can exist, the conditions provide a construction. The solutions for t=1,2 are also given in an alternate form useful for the study of trim arrays. The balanced solution for t=1 is more detailed than that known so far, and permits one to determine whether or not a solution exists in possibly fewer steps.     

On the construction of orthogonal F-squares of order n from an orthogonal array (n, k, s, 2) and an OL(s, t) set

Complete sets of orthogonal F-squares of order n = s^p, where g is a prime or prime power and p is a positive integer have been constructed by Hedayat, Raghavarao, and Seiden (1975). Federer (1977) has constructed complete sets of orthogonal F-squares of order n = 4t, where t is a positive integer. We give a general procedure for constructing orthogonal F-squares of order n from an orthogonal array (n, k, s, 2) and an OL(s, t) set, where n is not necessarily a prime or prime power. In particular, we show how to construct sets of orthogonal F-squares of order n = 2s^p, where s is a prime or prime power and p is a positive integer. These sets are shown to be near complete and approach complete sets as s and/or p become large. We have also shown how to construct orthogonal arrays by these methods. In addition, the best upper bound on the number t of orthogonal F(n, λ₁), F(n, λ₂), …, F(n, λ₁) squares is given.     

Some transitive Steiner triple systems of Bagchi and Bagchi

We show that the class of Steiner triple systems on 3^d points defined in Bagchi and Bagchi (J. Combin. Theory Ser. A 52 (1989) 51–61) closely resemble the systems defined through the designs of points and lines of an affine geometry of dimension d over F₃ in that they have a rich collection of hyperplanes and subspaces, all of which are designs of the same Bagchi-Bagchi type. The ternary codes and the automorphism groups of these designs can also be fully described.     

Concerning the vector v(m,t)

Let q = mt + 1 be a prime power, and let v(m, t) be the (m + 1)-vector (b₁, b₂, …, b_{m + 1}) of elements of GF(q) such that for each k, 1 ⩽ k ⩽ m + 1, the set {b_i−b_j:i∈{1,2,…m+1} − {m + 2 − k}, j  i + k(mod m + 2) and 1⩽j⩽m+1} forms a system of representatives for the cyclotomic classes of index m in GF(q). In this paper, we investigate the existence of such vectors. An upper bound on t for the existence of a v(m, t) is given for each fixed m unless both m and t are even, in which case there is no such a vector. Some special cases are also considered.     

Goodness-of-fit tests for semiparametric biased sampling models

Consider an s-sample biased sampling model in which the distribution function for each of the first s−1 samples is related to the unknown distribution function G of the sth sample by a known parametric selection bias weight function. Gilbert et al. (Biometrika 86 (1999) 27) gave a procedure for semiparametric maximum likelihood estimation of the parameters in this model. In many applications, information are scarce for basing the choice of the parametric weight function(s), motivating the need for goodness-of-fit tests of the hypothesis that the weight functions are correctly specified. Cramér–von Mises-type, Anderson–Darling-type, and Kolmogorov–Smirnov-type test statistics are studied which compare discrepancies between the empirical distribution of G and the semiparametric maximum likelihood estimator of G. Finite-sample properties of the tests are evaluated with simulations and with a real example of HIV genetic sequence data.     

FIRST-EXIT TIMES FOR POISSON SHOT NOISE

For a shot-noise process X(t) with Poisson arrival times and exponentially diminishing shocks of i.i.d. sizes, we consider the first time T _b at which a given level b > 0 is exceeded. An integral equation for the joint density of T _b and X(T _b ) is derived and, for the case of exponential jumps, solved explicitly in terms of Laplace transforms (LTs). In the general case we determine the ordinary LT of the function ? P(T _b > t) in terms of certain LTs derived from the distribution function H(x; t) = P(X(t) ≤ x), considered as a function of both variables x and t. Moreover, for G(t, u) = P(T _b > t, X(t) < u), that is the joint distribution function of sup_{0 ≤ s ≤ t} X(s) and X(t), an integro-differential equation is presented, whose unique solution is G(t, u).     

On block-transitive designs with affine automorphism group

Detailed necessary and sufficient conditions for a k-subset of AG(d, 3) to generate the block set of a block-transitive t-design with automorphism group AGL(d, 3) are derived for t = 3, 4, 5. Similar necessary conditions are found for the existence of a block-transitive design with automorphism group AGL(d, p) when p is an arbitrary odd prime. A search was carried out to find feasible parameter sets satisfying the implied divisibility conditions. The only ‘small’ feasible parameter sets found with k or v − k not exceeding 1000 were for t = 4 and (d, p) = (7, 3), (8, 3), and (3, 7). Examples of block-transitive 4-designs admitting AGL(7, 3) are found for each of the values k = 115, 116, 230, 437, and 552.     

The minimum coverage probability of confidence intervals in regression after a preliminary F test

Consider a linear regression model with regression parameter β=(β₁,…,β_p) and independent normal errors. Suppose the parameter of interest is θ=a^Tβ, where a is specified. Define the s-dimensional parameter vector τ=C^Tβ−t, where C and t are specified. Suppose that we carry out a preliminary F test of the null hypothesis H₀:τ=0 against the alternative hypothesis H₁:τ≠0. It is common statistical practice to then construct a confidence interval for θ with nominal coverage 1−α, using the same data, based on the assumption that the selected model had been given to us a priori (as the true model). We call this the naive 1−α confidence interval for θ. This assumption is false and it may lead to this confidence interval having minimum coverage probability far below 1−α, making it completely inadequate. We provide a new elegant method for computing the minimum coverage probability of this naive confidence interval, that works well irrespective of how large s is. A very important practical application of this method is to the analysis of covariance. In this context, τ can be defined so that H₀ expresses the hypothesis of “parallelism”. Applied statisticians commonly recommend carrying out a preliminary F test of this hypothesis. We illustrate the application of our method with a real-life analysis of covariance data set and a preliminary F test for “parallelism”. We show that the naive 0.95 confidence interval has minimum coverage probability 0.0846, showing that it is completely inadequate.     

On orthogonal arrays attaining Rao's bounds

Rao (1947) provided two inequalities on parameters of an orthogonal array OA(N,m,s,t). An orthogonal array attaining these Rao bounds is said to be complete. Noda (1979) characterized complete orthogonal arrays of t=4 (strength). We here investigate complete orthogonal arrays with s=2 (levels) and general t; and with t=2, 3 and general s.     

Families of Hadamard group divisible designs

By a family of designs we mean a set of designs whose parameters can be represented as functions of an auxiliary variable t where the design will exist for infinitely many values of t. The best known family is probably the family of finite projective planes with υ = b = t² + t + 1, r = k = t + 1, and λ = 1. In some instances, notably coding theory, the existence of families is essential to provide the degree of precision required which can well vary from one coding problem to another. A natural vehicle for developing binary codes is the class of Hadamard matrices. Bush (1977) introduced the idea of constructing semi-regular designs using Hadamard matrices whereas the present study is concerned mostly with construction of regular designs using Hadamard matrices. While codes constructed from these designs are not optimal in the usual sense, it is possible that they may still have substantial value since, with different values of λ₁ and λ₂, there are different error correcting capabilities.     

Limiting distributions of Kolmogorov-Lévy-type statistics under the alternative

Let X_i, 1 ≤ i ≤ n, be independent identically distributed random variables with a common distribution function F, and let G be a smooth distribution function. We derive the limit distribution of □{ρ_α(F_n, G) - _α(F, G)}, where F_n is the empirical distribution function based on X₁,…,X_n and _α is a Kolmogorov-Lévy-type metric between distribution functions. For α ≤ 0 and two distribution functions F and G the metric p_α is given by p_α(F, G) = inf {? ≤ 0: G(x - α?) - ? F(x) ≤ G(x + α?) + ? for all x ?}.     

Theory of factorial designs of the parallel flats type. I: the coefficient matrix

Let GF(s) be the finite field with s elements.(Thus, when s=3, the elements of GF(s) are 0, 1 and 2.)Let A(r×n), of rank r, and c_i(i=1,…,f), (r×1), be matrices over GF(s). (Thus, for n=4, r=2, f=2, we could have A=[¹¹¹⁰₀₁₂₁], c₁=[¹₀], c₂=[⁰₂].) Let T_i (i=1,…,f) be the flat in EG(n, s) consisting of the set of all the s^n?r solutions of the equations At=c_i, wheret′=(t₁,…,t_n) is a vector of variables.(Thus, EG(4, 3) consists of the 3⁴=81 points of the form (t₁,t₂,t₃,t₄), where t's take the values 0,1,2 (in GF(3)). The number of solutions of the equations At=c_i is s^n?r, where r=Rank(A), and the set of such solutions is said to form an (n?r)-flat, i.e. a flat of (n?r) dimensions. In our example, both T₁ and T₂ are 2-flats consisting of 3^4?2=9 points each. The flats T₁,T₂,…,T_f are said to be parallel since, clearly, no two of them can have a common point. In the example, the points of T₁ are (1000), (0011), (2022), (0102), (2110), (1121), (2201), (1212) and (0220). Also, T₂ consists of (0002), (2010), (1021), (2101), (1112), (0120), (1200), (0211) and (2222).) Let T be the fractional design for a sⁿ symmetric factorial experiment obtained by taking T₁,T₂,…,T_f together. (Thus, in the example, 3⁴=81 treatments of the 3⁴ factorial experiment correspond one-one with the points of EG(4,3), and T will be the design (i.e. a subset of the 81 treatments) consisting of the 18 points of T₁ and T₂ enumerated above.)In this paper, we lay the foundation of the general theory of such ‘parallel’ types of designs. We define certain functions of A called the alias component matrices, and use these to partition the coefficient matrix X (n×v), occuring in the corresponding linear model, into components X._j(j=0,1,…,g), such that the information matrix X is the direct sum of the X′._jX._j. Here, v is the total number of parameters, which consist of (possibly μ), and a (general) set of (geometric) factorial effects (each carrying (s?1) degrees of freedom as usual). For j≠0, we show that the spectrum of X′._jX._j does not change if we change (in a certain important way) the usual definition of the effects. Assuming that such change has been adopted, we consider the partition of the X._j into the X_ij (i=1,…,f). Furthermore, the X_ij are in turn partitioned into smaller matrices (which we shall here call the) X_ijh. We show that each X_ijh can be factored into a product of 3 matrices J, ζ (not depending on i,j, and h) and Q(j,h,i)where both the Kronecker and ordinary product are used. We introduce a ring R using the additive groups of the rational field and GF(s), and show that the Q(j,h,i) belong to a ring isomorphic to R. When s is a prime number, we show that R is the cyclotomic field. Finally, we show that the study of the X._j and X′._jX._j can be done in a much simpler manner, in terms of certain relatively small sized matrices over R.     

Orthogonal [k − 1, k + 1]-factorizations in graphs

Let G be a graph. Let F={F₁,F₂,...F_d} be a factorization of G and H be a subgraph of G. If H has exactly one edge in common with F_i for all i = 1,2,…,d, then we say that F is orthogonal to H. In this paper it is proved that for any d-matching M of a [kd − 1, kd + 1]-graphG, there is a [k − 1, k + 1]-factorization of G orthogonal to M where k ⩾ 2 is an integer.     

On the quantile process based on the autoregressive residuals

Let {X_t} be the stationary AR(p) process satisfying the difference equation X_t=β₁X_t−1 + … + β_pX_t−p+ε_t, where {ε_t} is a sequence of iid random variables with mean zero and finite variance. Motivated by a goodness of fit test on the true errors {ε_t}, we are led to study the asymptotic behavior of the quantile process based on residuals (the residual quantile process). Particularly, we concentrate on the deviations between the residual quantile process and the empirical process based on the true errors. In this asymptotic study, it is shown that the deviations converge to zero in probability uniformly over certain intervals with specific order as sample size increases. Here, these intervals are allowed to vary with the sample size n and converge to the unit interval as n goes to infinity. Then, based on our result and the strong approximation result of Csörgö and Révész (1978), we propose a goodness of fit test statistic of which limiting distribution is the same as of a functional form of a standard Brownian bridge.     

Mixed covering arrays of strength three with few factors

Covering arrays with mixed alphabet sizes, or mixed covering arrays, are useful generalizations of covering arrays that are motivated by software and network testing. Suppose that there are k factors, and that the ith factor takes values or levels from a set G_i of size g_i. A run is an assignment of an admissible level to each factor. A mixed covering array, MCA(N;t,k,g₁g₂…g_k), is a collection of N runs such that for any t distinct factors, i₁,i₂,…,i_t, every t-tuple from G_i1×G_i2×?×G_it occurs in factors i₁,i₂,…,i_t in at least one of the N runs. When g=g₁=g₂=?=g_k, an MCA(N;t,k,g₁g₂…g_k) is a CA(N;t,k,g). The mixed covering array number, denoted by MCAN(t,k,g₁g₂…g_k), is the minimum N for which an MCA(N;t,k,g₁g₂…g_k) exists. In this paper, we focus on the constructions of mixed covering arrays of strength three. The numbers MCAN(3,k,g₁g₂…g_k) are determined for all cases with k∈{3,4} and for most cases with k∈{5,6}.     

On a class of partially balanced incomplete block designs

Bose and Clatworthy (1955) showed that the parameters of a two-class balanced incomplete block design with λ1=1,λ2=0 and satisfying r <k can be expressed in terms of just three parameters r,k,t. Later Bose (1963) showed that such a design is a partial geometry (r,k,t). Bose, Shrikhande and Singhi (1976) have defined partial geometric designs (r,k,t,c), which reduce to partial geometries when c=0. In this note we prove that any two class partially balanced (PBIB) design with r <k, is a partial geometric design for suitably chosen r,k,t,c and express the parameters of the PBIB design in terms of r,k,t,c and λ2. We also show that such PBIB designs belong to the class of special partially balanced designs (SPBIB) studied by Bridges and Shrikhande (1974).     

Estimated Generalized Least Squares for a Heteroscedastic Regression Model

This paper considers the general linear regression model yc = X₁β+u_t under the heteroscedastic structure E(u_t) = 0, E(u²) =σ²- (X_tβ)², E(u_t u_s) = 0, tæs, t, s= 1, T. It is shown that any estimated GLS estimator for β is asymptotically equivalent to the GLS estimator under some regularity conditions. A three-step GLS estimator, which calls upon the assumption E(u_t²) =s?²(X,β)² for the estimation of the disturbance covariance matrix, is considered.