Showing papers in "Annals of the Institute of Statistical Mathematics in 1992"

Multinomial logistic regression algorithm

Abstract: The lower bound principle (introduced in Bohning and Lindsay 1988, Ann. Inst. Statist. Math., 40, 641–663), Bohning (1989, Biometrika, 76, 375–383) consists of replacing the second derivative matrix by a global lower bound in the Loewner ordering. This bound is used in the Newton-Raphson iteration instead of the Hessian matrix leading to a monotonically converging sequence of iterates. Here, we apply this principle to the multinomial logistic regression model, where it becomes specifically attractive.

Mean squared prediction error in the spatial linear model with estimated covariance parameters

Abstract: The problem considered is that of predicting the value of a linear functional of a random field when the parameter vector θ of the covariance function (or generalized covariance function) is unknown. The customary predictor when θ is unknown, which we call the EBLUP, is obtained by substituting an estimator Ĝj for θ in the expression for the best linear unbiased predictor (BLUP). Similarly, the customary estimator of the mean squared prediction error (MSPE) of the EBLUP is obtained by substituting Ĝj for θ in the expression f for the BLUP's MSPE; we call this the EMSPE. In this article, the appropriateness of the EMSPE as an estimator of the EBLUP's MSPE is examined, and alternative estimators of the EBLUP's MSPE for use when the EMSPE is inappropriate are suggested. Several illustrative examples show that the performance of the EMSPE depends on the strength of spatial correlation; the EMSPE is at its best when the spatial correlation is strong.

Estimating densities, quantiles, quantile densities and density quantiles

Abstract: To estimate the quantile density function (the derivative of the quantile function) by kernel means, there are two alternative approaches. One is the derivative of the kernel quantile estimator, the other is essentially the reciprocal of the kernel density estimator. We give ways in which the former method has certain advantages over the latter. Various closely related smoothing issues are also discussed.

A space-time clustering model for historical earthquakes

Abstract: This paper describes a generalization of Hawkes' self-exciting process in which each event creates a process of “offspring” with conditional intensity governed by a diffusion kernel. The process may be described as a space-time branching process with immigration, the immigration representing a background series of independent events. The model can be fitted by likelihood methods. As an illustration it is fitted to the catalogue of historical Italian earthquakes.

A form of multivariate gamma distribution

Abstract: Let V,, i = 1,...,k, be independent gamma random variables with shape ai, scale /3, and location parameter %, and consider the partial sums Z1 = V1, Z2 = 171 + V2, . • •, Zk = 171 +. • • + Vk. When the scale parameters are all equal, each partial sum is again distributed as gamma, and hence the joint distribution of the partial sums may be called a multivariate gamma. This distribution, whose marginals are positively correlated has several interesting properties and has potential applications in stochastic processes and reliability. In this paper we study this distribution as a multivariate extension of the three-parameter gamma and give several properties that relate to ratios and conditional distributions of partial sums. The general density, as well as special cases are considered.

Bayesian inference for the power law process

Abstract: The power law process has been used to model reliability growth, software reliability and the failure times of repairable systems. This article reviews and further develops Bayesian inference for such a process. The Bayesian approach provides a unified methodology for dealing with both time and failure truncated data. As well as looking at the posterior densities of the parameters of the power law process, inference for the expected number of failures and the probability of no failures in some given time interval is discussed. Aspects of the prediction problem are examined. The results are illustrated with two data examples.

Waiting time problems for a sequence of discrete random variables

Abstract: Let X1, X2,... be a sequence of nonnegative integer valued random variables.For each nonnegative integer i, we are given a positive integer ki. For every i = 0, 1, 2,..., Ei denotes the event that a run of i of length ki occurs in the sequence X1, X2,.... For the sequence X1, X2,..., the generalized pgf's of the distributions of the waiting times until the r-th occurrence among the events % MathType!MTEF!2!1!+-% feaafeart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXafv3ySLgzGmvETj2BSbqefm0B1jxALjhiov2D% aebbfv3ySLgzGueE0jxyaibaiiYdd9qrFfea0dXdf9vqai-hEir8Ve% ea0de9qq-hbrpepeea0db9q8as0-LqLs-Jirpepeea0-as0Fb9pgea% 0lrP0xe9Fve9Fve9qapdbaqaaeGacaGaaiaabeqaamaabaabcaGcba% WaaiWabeaacaWGfbWaaSbaaSqaaiaadMgaaeqaaaGccaGL7bGaayzF% aaWaa0baaSqaaiaadMgacqGH9aqpcaaIWaaabaGaeyOhIukaaaaa!43D8!\[\left\{ {E_i } \right\}_{i = 0}^\infty\]are obtained. Though our situations are general, the results are very simple. For the special cases that X's are i.i.d. and {0, 1}-valued, the corresponding results are consistent with previously published results.

General relations and identities for order statistics from non-independent non-identical variables

Abstract: Some recurrence relations and identities for order statistics are extended to the most general case where the random variables are assumed to be non-independent non-identically distributed. In addition, some new identities are given. The results can be used to reduce the computations considerably and also to establish some interesting combinatorial identities.

On exact D-optimal designs for regression models with correlated observations

Abstract: Let ~-* be an exact D-optimal design for a given regression model Y~ --- X~/~ + Z~. In this paper sufficient conditions are given for designing how the covariance matrix of Z~ may be changed so that not only ~-* remains D-optimal but also that the best linear unbiased estimator (BLUE) of ~ stays fixed for the design ~'*, although the covariance matrix of Z~. is changed. Hence under these conditions a best, according to D-optimality, BLUE of/~ is known for the model with the changed covariance matrix. The results may also be considered as determination of exact D-optimal designs for regression models with special correlated observations where the covariance matrices are not fully known. Various examples are given, especially for regression with intercept term, polynomial regression, and straight-line regression. A real example in electrocardiography is treated shortly.

Empirical Bayes with rates and best rates of convergence in u(x)C(θ) exp(-x/θ)-family: Estimation case

Abstract: Let % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaai4EaiaacI% cacaWGybWaaSbaaSqaaiaadMgaaeqaaOGaaiilaiabeI7aXnaaBaaa% leaacaWGPbaabeaakiaacMcacaGG9baaaa!3ED1!\[\{ (X_i ,\theta _i )\} \] be a sequence of independent random vectors where Xi, conditional on % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqiUde3aaS% baaSqaaiaadMgaaeqaaaaa!38BD!\[\theta _i \], has the probability density of the form % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamOzaiaacI% cacaWG4bGaaiiFaiabeI7aXnaaBaaaleaacaWGPbaabeaakiaacMca% cqGH9aqpcaWG1bGaaiikaiaadIhacaGGPaGaam4qaiaacIcacqaH4o% qCdaWgaaWcbaGaamyAaaqabaGccaGGPaGaaeyzaiaabIhacaqGWbGa% aiikaiabgkHiTiaadIhacaGGVaGaeqiUde3aaSbaaSqaaiaadMgaae% qaaOGaaiykaaaa!4FFF!\[f(x|\theta _i ) = u(x)C(\theta _i ){\text{exp}}( - x/\theta _i )\] and the unobservable % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqiUde3aaS% baaSqaaiaadMgaaeqaaaaa!38BD!\[\theta _i \] are i.i.d. according to an unknown G in some class G of prior distributions on Θ, a subset of % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaai4EaiabeI% 7aXjabg6da+iaaicdacaGG8bGaam4qaiaacIcacqaH4oqCcaGGPaGa% eyypa0JaaiikaiaadAgacaWG1bGaaiikaiaadIhacaGGPaGaaeyzai% aabIhacaqGWbGaaeikaiabgkHiTiaadIhacaGGVaGaeqiUdeNaaiyk% aiaadsgacaWG4bGaaiykamaaCaaaleqabaGaeyOeI0IaaGymaaaaki% abg6da+iaaicdacaGG9baaaa!54DE!\[\{ \theta > 0|C(\theta ) = (fu(x){\text{exp(}} - x/\theta )dx)^{ - 1} > 0\} \]. For a S(X1, ..., Xn, Xn+1)-measurable function % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqOXdy2aaS% baaSqaaiaad6gaaeqaaOGaaiilaaaa!397F!\[\phi _n ,\] let % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamOuamaaBa% aaleaacaWGUbaabeaakiabg2da9iaadweacaGGOaGaeqOXdy2aaSba% aSqaaiaad6gaaeqaaOGaeyOeI0IaeqiUde3aaSbaaSqaaiaad6gacq% GHRaWkcaaIXaaabeaakiaacMcadaahaaWcbeqaaiaaikdaaaaaaa!444A!\[R_n = E(\phi _n - \theta _{n + 1} )^2 \] denote the Bayes risk of % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqOXdy2aaS% baaSqaaiaad6gaaeqaaaaa!38C5!\[\phi _n \] and let R(G) denote the infimum Bayes risk with respect to G. For each integer s>1 we exhibit a class of S(X1, ..., Xn, Xn+1)-measurable functions % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqOXdy2aaS% baaSqaaiaad6gaaeqaaaaa!38C5!\[\phi _n \] such that for δ in [s−1, 1], % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaam4yamaaBa% aaleaacaaIWaaabeaakiaad6gadaahaaWcbeqaaiabgkHiTiaaikda% caWGZbGaai4laiaacIcacaaIXaGaey4kaSIaaGOmaiaadohacaGGPa% aaaOGaeyizImQaamOuamaaBaaaleaacaWGUbaabeaakiaacIcacqaH% gpGzdaWgaaWcbaGaamOBaaqabaGccaGGSaGaam4raiaacMcacqGHsi% slcaWGsbGaaiikaiaadEeacaGGPaGaeyizImQaam4yamaaBaaaleaa% caaIXaaabeaakiaad6gadaahaaWcbeqaaiabgkHiTiaaikdacaGGOa% Gaam4Caiabes7aKjabgkHiTiaaigdacaGGPaGaai4laiaacIcacaaI% XaGaey4kaSIaaGOmaiaadohacaGGPaaaaaaa!5F94!\[c_0 n^{ - 2s/(1 + 2s)} \leqslant R_n (\phi _n ,G) - R(G) \leqslant c_1 n^{ - 2(s\delta - 1)/(1 + 2s)} \] under certain conditions on u and G. No assumptions on the form or smoothness of u is made, however. Examples of functions u, including one with infinitely many discontinuities, are given for which our conditions reduce to some moment conditions on G. When Θ is bounded, for each integer s>1 S(X1, ..., Xn, Xn+1)-measurable functions % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqOXdy2aaS% baaSqaaiaad6gaaeqaaaaa!38C5!\[\phi _n \] are exhibited such that for δ in % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaai4waiaaik% dacaGGVaGaam4CaiaacYcacaaIXaGaaiyxaiaadogadaqhaaWcbaGa% aGimaaqaaiaacEcaaaGccaWGUbWaaWbaaSqabeaacqGHsislcaaIYa% Gaam4Caiaac+cacaGGOaGaaGymaiabgUcaRiaaikdacaWGZbGaaiyk% aaaakiabgsMiJkaadkfadaWgaaWcbaGaamOBaaqabaGccaGGOaGaeq% OXdy2aaSbaaSqaaiaad6gaaeqaaOGaaiilaiaadEeacaGGPaGaeyOe% I0IaamOuaiaacIcacaWGhbGaaiykaiabgsMiJkaadogadaqhaaWcba% GaaGymaaqaaiaacEcaaaGccaWGUbWaaWbaaSqabeaacqGHsislcaaI% YaGaam4Caiabes7aKjaac+cacaGGOaGaaGymaiabgUcaRiaaikdaca% WGZbGaaiykaaaaaaa!637D!\[[2/s,1]c_0^' n^{ - 2s/(1 + 2s)} \leqslant R_n (\phi _n ,G) - R(G) \leqslant c_1^' n^{ - 2s\delta /(1 + 2s)} \]. Examples of functions u and class g are given where the above lower and upper bounds are achieved.

Integrated squared error of kernel-type estimator of distribution function

Abstract: Let X 1 ,...,X n be a random sample drawn from distribution function F(x) with density function f(x) and suppose we want to estimate X(x). It is already shown that kernel estimator of F(x) is better than usual empirical distribution function in the sense of mean integrated squared error. In this paper we derive integrated squared error of kernel estimator and compare the error with that of the empirical distribution function. It is shown that the superiority of kernel estimators is not necessarily true in the sense of integrated squared error.

Approximations to the birthday problem with unequal occurrence probabilities and their application to the surname problem in Japan

Abstract: Let X1, X2,..., Xn be iid random variables with a discrete distribution {pi}i=1m. We will discuss the coincidence probability Rn, i.e., the probability that there are members of {Xi} having the same value. If m=365 and pi≡1/365, this is the famous birthday problem. Also we will give two kinds of approximation to this probability. Finally we will give two applications. The first is the estimation of the coincidence probability of surnames in Japan. For this purpose, we will fit a generalized zeta distribution to a frequency data of surnames in Japan. The second is the true birthday problem, that is, we will evaluate the birthday probability in Japan using the actual (non-uniform) distribution of birthdays in Japan.

Large sample properties of estimates of a discrete grade of membership model

Abstract: Increasingly, fuzzy partitions are being used in multivariate clas- sification problems as an alternative to the crisp classification procedures com- monly used. One such fuzzy partition, the grade of membership model, parti- tions individuals into fuzzy sets using multivariate categorical data. Although the statistical methods used to estimate fuzzy membership for this model are based on maximum likelihood methods, large sample properties of the estima- tion procedure are problematic for two reasons. First, the number of incidental parameters increases with the size of the sample. Second, estimated param- eters fall on the boundary of the parameter space with non-zero probability. This paper examines the consistency of the likelihood approach when estimat- ing the components of a particular probability model that gives rise to a fuzzy partition. The results of the consistency proof are used to determine the large sample distribution of the estimates. Common methods of classifying individ- uals based on multivariate observations attempt to place each individual into crisply defined sets. The fuzzy partition allows for individual to individual heterogeneity, beyond simply errors in measurement, by defining a set of pure type characteristics and determining each individual's distance from these pure types. Both the profiles of the pure types and the heterogeneity of the indi- viduals must be estimated from data. These estimates empirically define the fuzzy partition. In the current paper, this data is assumed to be categorical data. Because of the large number of parameters to be estimated and the limitations of categorical data, one may be concerned about whether or not the fuzzy partition can be estimated consistently. This paper shows that if heterogeneity is measured with respect to a fixed number of moments of the grade of membership scores of each individual, the estimated fuzzy partition is consistent.

Treating bias as variance for experimental design purposes

Abstract: When an empirical model is fitted to data, bias can arise from terms that have not been incorporated, and this can have an important effect on the choice of an experimental design. Here, the biases are treated as random, and the consequences of this action are explored for the fitting of models of first and second order.

Consecutive k-out-of-n systems with maintenance

Abstract: A consecutive k-out-of-n system consists of n linearly or cycli- cally ordered components such that the system fails if and only if at least k consecutive components fail. In this paper we consider a maintained system where each component is repaired independently of the others according to an exponential distribution. Assuming general lifetime distributions for system's components we prove a limit theorem for the time to first failure of both linear and circular systems.

Further developments on some dependence orderings for continuous bivariate distributions

Abstract: The dependence orderings, “more associated” and “more regression dependent”, due to Schriever (1986, Order Dependence, Centre for Mathematics and Computer Sciences, Amsterdam; 1987, Ann. Statist., 15, 1208–1214) and Yanagimoto and Okamoto (1969, Ann. Inst. Statist. Math., 21, 489–505) respectively, are studied in detail for continuous bivariate distributions. Equivalent forms of the orderings under some conditions are given so that the orderings are more easily checkable for some bivariate distributions. For several parametric bivariate families, the dependence orderings are shown to be equivalent to an ordering of the parameter. A study of functionals that are increasing with respect to the “more associated ordering” leads to inequalities, measures of dependence as well as a way of checking that this ordering does not hold for two distributions.

Expansions for statistics involving the mean absolute deviations

Abstract: Expansion for the difference of mean absolute deviations from the sample mean and the population mean is derived. This result is used to obtain strong representations for mean absolute deviations from the sample mean and the sample median. Edgeworth expansions for some scale invariant statistics involving the mean absolute deviations are studied. These expansions are shown to be valid in spite of the presence of a lattice variable.

Average worth and simultaneous estimation of the selected subset

Abstract: Suppose a subset of populations is selected from the given k gamma G(θ i,p ) (i = 1,2,...,k)populations, using Gupta's rule (1963, Ann. Inst. Statist. Math., 14, 199–216). The problem of estimating the average worth of the selected subset is first considered. The natural estimator is shown to be positively biased and the UMVUE is obtained using Robbins' UV method of estimation (1988, Statistical Decision Theory and Related Topics IV, Vol. 1 (eds. S. S. Gupta and J. O. Berger), 265–270, Springer, New York). A class of estimators that dominate the natural estimator for an arbitrary k is derived. Similar results are observed for the simultaneous estimation of the selected subset.

Optimum bandwidths and kernels for estimating certain discontinuous densities

Abstract: Rosenblatt and Parzen proposed a well-known estimator fn for an unknown density function f, and later Schuster suggested a modification % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaybyaeqale% qabaGaaiOxaaqdbaGaamOzaaaaaaa!3851!\[\mathop f\limits^\^ \]n to rectify certain drawbacks of fn. This paper gives the asymptotically optimum bandwidth and kernel for % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaybyaeqale% qabaGaaiOxaaqdbaGaamOzaaaaaaa!3851!\[\mathop f\limits^\^ \]n under the standard measure of IMSE when f is discontinuous at one or both endpoints of its support. We also consider an alternative definition of the IMSE under which the optimum bandwidths and kernels for % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaybyaeqale% qabaGaaiOxaaqdbaGaamOzaaaaaaa!3851!\[\mathop f\limits^\^ \]n and fn are derived. The latter supplement van Eeden's results.

On Laplace continued fraction for the normal integral

Abstract: The Laplace continued fraction is derived through a power series. It provides both upper bounds and lower bounds of the normal tail probability % MathType!MTEF!2!1!+-% feaafeart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXafv3ySLgzGmvETj2BSbqefm0B1jxALjhiov2D% aebbfv3ySLgzGueE0jxyaibaiGc9yrFr0xXdbba91rFfpec8Eeeu0x% Xdbba9frFj0-OqFfea0dXdd9vqaq-JfrVkFHe9pgea0dXdar-Jb9hs% 0dXdbPYxe9vr0-vr0-vqpWqaaeaabiGaciaacaqabeaadaqaaqGaaO% qaaiqbfA6agzaaraaaaa!3DC0!\[\bar \Phi\](x), it is simple, it converges for x>0, and it is by far the best approximation for x≥3. The Laplace continued fraction is rederived as an extreme case of admissible bounds of the Mills' ratio, % MathType!MTEF!2!1!+-% feaafeart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXafv3ySLgzGmvETj2BSbqefm0B1jxALjhiov2D% aebbfv3ySLgzGueE0jxyaibaiGc9yrFr0xXdbba91rFfpec8Eeeu0x% Xdbba9frFj0-OqFfea0dXdd9vqaq-JfrVkFHe9pgea0dXdar-Jb9hs% 0dXdbPYxe9vr0-vr0-vqpWqaaeaabiGaciaacaqabeaadaqaaqGaaO% qaaiqbfA6agzaaraaaaa!3DC0!\[\bar \Phi\](x)/ϕ(x), in the family of ratios of two polynomials subject to a monotone decreasing absolute error. However, it is not optimal at any finite x. Convergence at the origin and local optimality of a subclass of admissible bounds are investigated. A modified continued fraction is proposed. It is the sharpest tail bound of the Mills' ratio, it has a satisfactory convergence rate for x≥1 and it is recommended for the entire range of x if a maximum absolute error of 10-4 is required.

Minimum f-divergence estimators and quasi-likelihood functions

Abstract: Maximum quasi-likelihood estimators have several nice asymptotic properties. We show that, in many situations, a family of estimators, called the minimum f-divergence estimators, can be defined such that each estimator has the same asymptotic properties as the maximum quasi-likelihood estimator. The family of minimum f-divergence estimators include the maximum quasi-likelihood estimators as a special case. When a quasi-likelihood is the log likelihood from some exponential family, Amari's dual geometries can be used to study the maximum likelihood estimator. A dual geometric structure can also be defined for more general quasi-likelihood functions as well as for the larger family of minimum f-divergence estimators. The relationship between the f-divergence and the quasi-likelihood function and the relationship between the f-divergence and the power divergence is discussed.

Asymptotic risk behavior of mean vector and variance estimators and the problem of positive normal mean

Abstract: Asymptotic risk behavior of estimators of the unknow variance and of the unknown mean vector in a multivariate normal distribution is considered for a general loss. It is shown that in both problems this characteristic is related to the risk in an estimation problem of a positive normal mean under quadratic loss function. A curious property of the Brewster-Zidek variance estimator of the normal variance is also noticed.

Efficiency of connected binary block designs when a single observation is unavailable

Abstract: In this paper the problem of finding the design efficiency is considered when a single observation is unavailable in a connected binary block design. The explicit expression of efficiency is found for the resulting design when the original design is a balanced incomplete block design or a group divisible, singular or semiregular or regular with λ1>0, design. The efficiency does not depend on the position of the unavailable observation. For a regular group divisible design with λ1>0, the efficiency depends on the position of the unavailable observation. The bounds, both lower and upper, on the efficiency are given in this situation. The efficiencies of designs resulting from a balanced incomplete block design and a group divisible design are in fact high when a single observation is unavailable.

Jackknifing in generalized linear models

Abstract: In a generalized linear model, the jackknife estimator of the asymptotic covariance matrix of the maximum likelihood estimator is shown to be consistent. The corresponding jackknife studentized statistic is asymptotically normal. In addition, these results remain true even if there exist unequal dispersion parameters in the model. On the other hand, the variance estimator and the studentized statistic based on the standard method (substitution and linearization) do not enjoy this robustness property against the presence of unequal dispersion parameters.

Some characterizations of the uniform distribution with applications to random number generation

Abstract: Let U and V be independent random variables with continuous density function on the interval (0, 1). We describe families of functions g for which uniformity of U and V is equivalent to uniformity of g(U, V) on (0, 1). These results axe used to prescribe methods for improving the quality of pseudo-random number generators by making them closer in distribution to the U(0, 1) distribution.

Limit theorems for the minimum interpoint distance between any pair of i.i.d. random points in Rd

Abstract: The limit theorem for the minimum interpoint distance between any pair of i.i.d. random points in R d with common density f∈L2 was studied by a method which makes use of the convergence of point processes. Some one-dimensional examples with f∉L2 (including the cases Beta and Gamma distributions) were also considered.

Subsample and half-sample methods

Abstract: Hartigan's subsample and half-sample methods are both shown to be inefficient methods of estimating the sampling distributions. In the sample mean case the bootstrap is known to correct for skewness. But irrespective of the population, the estimates based on the subsample method, have skewness factor zero. This problem persists even if we take only samples of size less than or equal to half of the original sample. For linear statistics it is possible to correct this by considering estimates based on subsamples of size νn, when the sample size is n. In the sample mean case ν can be taken as % MathType!MTEF!2!1!+-% feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaeimaiaab6% cacaqG1aGaaeikaiaabgdacaqGGaGaeyOeI0IaaeiiaiaabgdacaqG% VaWaaOaaaeaacaqG1aaaleqaaOGaaeykaaaa!3E8A!\[{\text{0}}{\text{.5(1 }} - {\text{ 1/}}\sqrt {\text{5}} {\text{)}}\]. In spite of these negative results, the half-sample method is useful in estimating the variance of sample quantiles. It is shown that this method gives as good an estimate as that given by the bootstrap method. A major advantage of the half-sample method is that it is shown to be robust in estimating the mean square error of estimators of parameters of a linear regression model when the errors are heterogeneous. Bootstrap is known to give inconsistent results in this case; although, it is more efficient in the case of homogeneous errors.

On bilinear forms in normal variables

Abstract: Bilinear forms in normal variables when the matrices of the forms are rectangular are considered. Explicit expressions for the cumulants, joint cumulants and joint cumulants of bilinear and quadratic forms are given. Necessary and sufficient conditions are established for the independence of two bilinear forms as well as a bilinear and a quadratic form. Special cases are shown to agree with known results.

Bayesian priors based on a parameter transformation using the distribution function

Abstract: One of the tasks of the Bayesian consulting statistician is to elicit prior information from his client who may be unfamiliar with parametric statistical models. In some cases it may be more illuminating to base a prior distribution for parameter θ on the transformed version F(⋎/θ), where F is the data distribution function and v is a designated reference value, rather than on θ directly. This approach is outlined and explored in various directions to assess its implications. Some applications are given, including general linear regression and transformed linear models.

Minimax invariant estimator of a continuous distribution function

Abstract: Consider the problems of the continuous invariant estimation of a distribution function with a wide class of loss functions. It has been conjectured for long that the best invariant estimator is minimax for all sample sizes n≥1. This conjecture is proved in this short note.