On characterizing distributions via regression of functions of generalized order statistics

Let X(1,n,m₁,k),X(2,n,m₂,k),…,X(n,n,m,k) be n generalized order statistics from a continuous distribution F which is strictly increasing over (a,b),−∞≤a<b≤∞, the support of F. Let g be an absolutely continuous and monotonically increasing function in (a,b) with finite g(a⁺),g(b⁻) and E(g(X)). Then for some positive integer s,1<s≤n, we give characterization of distributions by means of

     

Sequential experimental design and response optimisation

We consider the situation where one wants to maximise a functionf(θ,x) with respect tox, with θ unknown and estimated from observationsy _k . This may correspond to the case of a regression model, where one observesy _k =f(θ,x _k )+ε _k , with ε _k some random error, or to the Bernoulli case wherey _k ∈{0, 1}, with Pr[y _k =1|θ,x _k |=f(θ,x _k ). Special attention is given to sequences given by , with an estimated value of θ obtained from (x₁, y₁),...,(x _k ,y _k ) andd _k (x) a penalty for poor estimation. Approximately optimal rules are suggested in the linear regression case with a finite horizon, where one wants to maximize ∑ _i=1 ^N w _i f(θ, x _i ) with {w _i } a weighting sequence. Various examples are presented, with a comparison with a Polya urn design and an up-and-down method for a binary response problem.     

Reliabilities for (n,f, k (i,j)) and ⟨n,f, k (i,j)⟩ Systems

The (n,f,k(i,j)):F(? n,f,k(i,j)?:F) system consists of n components ordered in a line or circle, while the system fails if, and only if, there exist at least f failed components OR (AND) at least k consecutive failed components among components i,i + 1,…,j ? 1,j. In this article, we present the system reliability formulae for these systems with product of matrices by means of a two-stage finite Markov chain imbedding approach, a technique first used by Cui et al. (2002 Cui , L. R. , Kuo , W. , Xie , M. ( 2002 ). On (f,g)-out-of-((i,j),n) systems and its reliability . In: Third International Conference on Mathematical Methods in Reliability Methodology and Practice , June 17–20 , Norway , Trondheim , pp. 173 – 176 . [Google Scholar]). In addition, their dual systems, denoted by (n,f,k(i,j)):G and ? n,f,k(i,j)?:G, are also introduced. Two numerical examples are given to illustrate the results.     

Pseudo New Modified Latin Square (NML m (m)) Type PBIB Designs

A New Modified Latin square [NML _i (m)] association scheme with i constraints for v = m ² treatments was introduced by Garg (2008 Garg , D. K. ( 2008 ). New modified Latin square (NMLi) type PBIB designs . J. Math. Syst. Sci. 1 ( 4 ): 83 – 89 . [Google Scholar]). In this article, a new association scheme known as Pseudo New Modified Latin square [Pseudo NML _m (m)] type association scheme is defined. The parameters of Pseudo NML _m (m) association scheme turned out to be parameters of NML _i (m) association scheme by taking i = m in NML _i (m) association scheme. The Pseudo NML _m (m) association scheme will be the usual NML _m (m) association scheme when m is a prime or a prime power. The PBIB designs following Pseudo NML _m (m) association scheme will be called the Pseudo NML _m (m) type PBIB designs. Analysis of Pseudo NML _m (m) designs along with a construction method of these designs is also given in this article.     

Coupling methods in approximations

Let X and Y be two arbitrary k-dimensional discrete random vectors, for k ≥ 1. We prove that there exists a coupling method which minimizes P( X ≠ Y ). This result is used to find the least upper bound for the metric d( X, Y ) = sup_A|P( X ∈ A ) ? P( Y ∈ A )| and to derive the inequality d(Σ X _i, Σ Y _i) ≤ Σd( X _i, Y _i). We thus obtain a unified method to measure the disparity between the distributions of sums of independent random vectors. Several examples are given.     

Simultaneous Canonization of Linear Models

Canonical form plays a similar role in linear models to spectral decomposition in matrix analysis. Let X = (X ₁,…, X _n )′ be a random vector with expectation Aβ and the variance–covariance matrix σV, where V is positive definite and let rank(A) = r. Then there exists a nonsingular linear transformation from X to T = (T ₁,…, T _n )′, such that ET _i  = η _i , for i = 1,…, r and zero for i > r, while cov(T _i , T _j ) = δ _ij σ. This canonical form, introduced by Ko?odziejczyk (1935 Ko?odziejczyk , S. ( 1935 ). On an important class of statistical hypotheses . Biometrika 27 : 161 – 190 .[Crossref] , [Google Scholar]), was used, among others, by Scheffé (1959 Scheffé , H. ( 1959 ). Analysis of Variance . New York : Wiley . [Google Scholar]) and by Lehmann (1959, 1986 Lehmann , E. L. (1959, 1986 ). Testing Statistical Hypotheses . New York : Wiley . [Google Scholar]). This technique is extended here for arbitrary (possibly singular) V and for simultaneous canonization of two models of this type.     

L models and multiple regressions designs

Given an orthogonal model

${{\bf \lambda}}=\sum_{i=1}^m{{{\bf X}}_i}{\boldsymbol{\alpha}}_i$

an L model

${{\bf y}}={\bf L}\left(\sum_{i=1}^m{{{\bf X}}_i}{\boldsymbol{\alpha}}_i\right)+{\bf e}$

is obtained, and the only restriction is the linear independency of the column vectors of matrix L. Special cases of the L models correspond to blockwise diagonal matrices L = D(L ₁, . . . , L _c ). In multiple regression designs this matrix will be of the form

${\bf L}={\bf D}(\check{{\bf X}}_1,\ldots,\check{{\bf X}}_{c})$

with \({\check{{\bf X}}_j, j=1,\ldots,c}\) the model matrices of the individual regressions, while the original model will have fixed effects. In this way, we overcome the usual restriction of requiring all regressions to have the same model matrix.

     

Estimating parameters of a selected Pareto population

Let Π₁,…,Π_k be k populations with Π_i being Pareto with unknown scale parameter α_i and known shape parameter β_i;i=1,…,k. Suppose independent random samples (X_i1,…,X_in), i=1,…,k of equal size are drawn from each of k populations and let X_i denote the smallest observation of the ith sample. The population corresponding to the largest X_i is selected. We consider the problem of estimating the scale parameter of the selected population and obtain the uniformly minimum variance unbiased estimator (UMVUE) when the shape parameters are assumed to be equal. An admissible class of linear estimators is derived. Further, a general inadmissibility result for the scale equivariant estimators is proved.     

Is the Selected Population the Best?—Location and Scale Parameter Cases

We observe X ₁,…,X _k , where X _i has density f(x,θ _i ) possessing monotone likelihood ratio. The best population corresponds to the largest θ _i . We select the population corresponding to the largest X _i . The goal is to attach the best possible p-value to the inference: the selected population has the uniquely largest θ _i . Gutmann and Maymin (1987 Gutmann , S. , Maymin , Z. ( 1987 ). Is the selected population the best? Ann. Statist . 15 : 456 – 461 .[Crossref], [Web of Science ®] , [Google Scholar]) considered the location parameter case and derived the supremum of the error probability by conditioning on S, the index of the largest X _i . Using this conditioning approach, Kannan and Panchapakesan (2009 Kannan , N. , Panchapakesan , S. ( 2009 ). Does the selected normal population have the smallest variance? Amer. J. Math. Management Sci . 29 : To appear . [Google Scholar]) considered the problem for the gamma family. We consider here a unified approach to both the location and scale parameter cases, and obtain the supremum of the error probability without using conditioning.     

On multivariate variable-kernel density estimates for time series

Let X₁ be a strictly stationary multiple time series with values in R^d and with a common density f. Let X₁,.,.,X_n, be n consecutive observations of X₁. Let k = k_n, be a sequence of positive integers, and let H_ni be the distance from X_i to its kth nearest neighbour among X_j, j i. The multivariate variable-kernel estimate f_n, of f is defined by where K is a given density. The complete convergence of f_n, to f on compact sets is established for time series satisfying a dependence condition (referred to as the strong mixing condition in the locally transitive sense) weaker than the strong mixing condition. Appropriate choices of k are explicitly given. The results apply to autoregressive processes and bilinear time-series models.