THE OPTIMAL CLASSIFICATION RULE FOR EXPONENTIAL POPULATIONS

Let π₁ and π₂ be two exponential populations with location parameters α and β respectively. It is shown that the probability of misclassification for the optimal rule, R_E, depends on α and β only through their ratio α/β=r, and that the smaller the ratio r (i.e. the smaller α is compared to β), the greater the superiority of the optimal rule R_E over the commonly used rule R_N. When α and β are unknown, the sample-based version R_E(s) of R_E exhibits the same pattern of superiority over the sample-based R_N(s) of R_N.     

Robustness of the linear discriminant function to nonnormality: Edgeworth series distribution

The effects of applying the normal classificatory rule to a nonnormal population are studied here. These are assessed through the distribution of the misclassification errors in the case of the Edgeworth type distribution. Both theoretical and empirical results are presented. An examination of the latter shows that the effects of this type of nonnormality are marginal. The probability of misclassification of an observation from ∏₁, using the appropriate LR rule, is always larger than one using the normal approximation (μ₁<μ₂). Converse condition holds for the misclassification of an observation from ∏₂. Overall error rates are not affected by the skewness factor to any great extent.     

Sequential Fixed-Width Confidence Interval for a Function of the Parameters

Let X₁…, X_m and Y₁…, Y_n be two independent sequences of i.i.d. random variables with distribution functions F_x(.|θ) and F_y(. | φ) respectively. Let g(θ, φ) be a real-valued function of the unknown parameters θ and φ. The purpose of this paper is to suggest a sequential procedure which gives a fixed-width confidence interval for g(θ, φ) so that the coverage probability is approximately α (preas-signed). Certain asymptotic optimality properties of the sequential procedure are established. A Monte Carlo study is presented.     

Bayesian prediction of the number of change points in the piecewise linear model

Summary The problem of predicting the number of change points in a piecewise linear model is studied from a Bayesian viewpoint. For a given a priori joint probability functionf _R,C=f_Rf _C/R, whereR is the number of change points andC=C′(R)=(C₁,…,C_R) is the change-point epoch vector, the marginal posterior probability functionf _R.C/Y is obtained, and then used to find predictors forR andC(R).     

A Restricted Subset Selection Rule for Selecting at Least One of the t Best Normal Populations in Terms of Their Means When Their Common Variance is Known,Case II

Consider k( ? 2) normal populations with unknown means μ₁, …, μ_k, and a common known variance σ². Let μ_[1] ? ??? ? μ_[k] denote the ordered μ_i.The populations associated with the t(1 ? t ? k ? 1) largest means are called the t best populations. Hsu and Panchapakesan (2004) proposed and investigated a procedure R_HPfor selecting a non empty subset of the k populations whose size is at most m(1 ? m ? k ? t) so that at least one of the t best populations is included in the selected subset with a minimum guaranteed probability P* whenever μ_{[k ? t + 1]} ? μ_{[k ? t]} ? δ*, where P*?and?δ* are specified in advance of the experiment. This probability requirement is known as the indifference-zone probability requirement. In the present article, we investigate the same procedure R_HP for the same goal as before but when k ? t < m ? k ? 1 so that at least one of the t best populations is included in the selected subset with a minimum guaranteed probability P* whatever be the configuration of the unknown μ_i. The probability requirement in this latter case is termed the subset selection probability requirement. Santner (1976) proposed and investigated a different procedure (R_S) based on samples of size n from each of the populations, considering both cases, 1 ? m ? k ? t and k ? t < m ? k. The special case of t = 1 was earlier studied by Gupta and Santner (1973) and Hsu and Panchapakesan (2002) for their respective procedures.     

Confidence Intervals for Nonparametric Regression Functions with Missing Data

Suppose that we have a nonparametric regression model Y = m(X) + ε with X ∈ R^p, where X is a random design variable and is observed completely, and Y is the response variable and some Y-values are missing at random. Based on the “complete” data sets for Y after nonaprametric regression imputation and inverse probability weighted imputation, two estimators of the regression function m(x₀) for fixed x₀ ∈ R^p are proposed. Asymptotic normality of two estimators is established, which is used to construct normal approximation-based confidence intervals for m(x₀). We also construct an empirical likelihood (EL) statistic for m(x₀) with limiting distribution of χ²₁, which is used to construct an EL confidence interval for m(x₀).     

Selection rules based on divergences

This paper deals with a special adaptive estimation problem, namely how can one select for each set of i.i.d. data X ₁, …, X _n the better of two given estimates of the data-generating probability density. Such a problem was studied by Devroye and Lugosi [Combinatorial Methods in Density Estimation, Springer, Berlin, 2001] who proposed a feasible suboptimal selection (called the Scheffé selection) as an alternative to the optimal but nonfeasible selection which minimizes the L₁-error. In many typical situations, the L₁-error of the Scheffé selection was shown to tend to zero for n→∞ as fast as the L₁-error of the optimal estimate. This asymptotic result was based on an inequality between the total variation errors of the Scheffé and optimal selections. The present paper extends this inequality to the class of φ-divergence errors containing the L₁-error as a special case. The first extension compares the φ-divergence errors of the mentioned Scheffé and optimal selections of Devroye and Lugosi. The second extension deals with a class of generalized Scheffé selections adapted to the φ-divergence error criteria and reducing to the classical Scheffé selection for the L₁-criterion. It compares the φ-divergence errors of these feasible selections and the optimal nonfeasible selections minimizing the φ-divergence errors. Both extensions are motivated and illustrated by examples.     

Mutiple three-decision procedure for comparing several exponential treatments with the best control

In this paper, the three-decision procedures to classify p treatments as better than or worse than one control, proposed for normal/symmetric probability models [Bohrer, Multiple three-decision rules for parametric signs. J. Amer. Statist. Assoc. 74 (1979), pp. 432–437; Bohrer et al., Multiple three-decision rules for factorial simple effects: Bonferroni wins again!, J. Amer. Statist. Assoc. 76 (1981), pp. 119–124; Liu, A multiple three-decision procedure for comparing several treatments with a control, Austral. J. Statist. 39 (1997), pp. 79–92 and Singh and Mishra, Classifying logistic populations using sample medians, J. Statist. Plann. Inference 137 (2007), pp. 1647–1657]; in the literature, have been extended to asymmetric two-parameter exponential probability models to classify p(p≥1) treatments as better than or worse than the best of q(q≥1) control treatments in terms of location parameters. Critical constants required for the implementation of the proposed procedure are tabulated for some pre-specified values of probability of no misclassification. Power function of the proposed procedure is also defined and a common sample size necessary to guarantee various pre-specified power levels are tabulated. Optimal allocation scheme is also discussed. Finally, the implementation of the proposed methodology is demonstrated through numerical example.     

The application of bias to discriminant analysis

When classification rules are constructed using sample estimatest it is known that the probability of misclassification is not minimized. This article introduces a biased minimum X² rule to classify items from a multivariate normal population. Using the principle of variance reduction, the probability of misclassification is reduced when the biased procedure is employed. Results of sampling experiments over a broad range of conditions are provided to demonstrate this improvement.     

The Target Parameter of Adjusted R-Squared in Fixed-Design Experiments

R-squared (R²) and adjusted R-squared (R²_Adj) are sometimes viewed as statistics detached from any target parameter, and sometimes as estimators for the population multiple correlation. The latter interpretation is meaningful only if the explanatory variables are random. This article proposes an alternative perspective for the case where the x’s are fixed. A new parameter is defined, in a similar fashion to the construction of R², but relying on the true parameters rather than their estimates. (The parameter definition includes also the fixed x values.) This parameter is referred to as the “parametric” coefficient of determination, and denoted by ρ²_*. The proposed ρ²_* remains stable when irrelevant variables are removed (or added), unlike the unadjusted R², which always goes up when variables, either relevant or not, are added to the model (and goes down when they are removed). The value of the traditional R²_Adj may go up or down with added (or removed) variables, either relevant or not. It is shown that the unadjusted R² overestimates ρ²_*, while the traditional R²_Adj underestimates it. It is also shown that for simple linear regression the magnitude of the bias of R²_Adj can be as high as the bias of the unadjusted R² (while their signs are opposite). Asymptotic convergence in probability of R²_Adj to ρ²_* is demonstrated. The effects of model parameters on the bias of R² and R²_Adj are characterized analytically and numerically. An alternative bi-adjusted estimator is presented and evaluated.     

Bayesian Estimation of Change Point in Inverse Weibull Sequence

A sequence of independent lifetimes X ₁,…, X _m , X _m+1,…, X _n were observed from inverse Weibull distribution with mean stress θ₁ and reliability R ₁(t ₀) at time t ₀ but later it was found that there was a change in the system at some point of time m and it is reflected in the sequence after X _m by change in mean stress θ₁ and in reliability R ₂(t ₀) at time t ₀. The Bayes estimators of m, R ₁(t ₀) and R ₂(t ₀) are derived when a poor and a more detailed prior information is introduced into the inferential procedure. The effects of correct and wrong prior information on the Bayes estimators are studied.     

Limiting behavior of randomly weighted averages of symmetric heavy-tailed random variables

In this paper we consider a sequence of independent continuous symmetric random variables X₁, X₂, …, with heavy-tailed distributions. Then we focus on limiting behavior of randomly weighted averages S_n = R⁽ⁿ⁾₁X₁ + ??? + R⁽ⁿ⁾_nX_n, where the random weights R⁽ⁿ⁾₁, …, R_n⁽ⁿ⁾ which are independent of X₁, X₂, …, X_n, are the cuts of (0, 1) by the n ? 1 order statistics from a uniform distribution. Indeed we prove that c_nS_n converges in distribution to a symmetric α-stable random variable with c_n = n^{1 ? 1/α}/Γ^1/α(α + 1).     

Exact average coverage probabilities and confidence coefficients of confidence intervals for discrete distributions

For a confidence interval (L(X),U(X)) of a parameter θ in one-parameter discrete distributions, the coverage probability is a variable function of θ. The confidence coefficient is the infimum of the coverage probabilities, inf  _θ P _θ (θ∈(L(X),U(X))). Since we do not know which point in the parameter space the infimum coverage probability occurs at, the exact confidence coefficients are unknown. Beside confidence coefficients, evaluation of a confidence intervals can be based on the average coverage probability. Usually, the exact average probability is also unknown and it was approximated by taking the mean of the coverage probabilities at some randomly chosen points in the parameter space. In this article, methodologies for computing the exact average coverage probabilities as well as the exact confidence coefficients of confidence intervals for one-parameter discrete distributions are proposed. With these methodologies, both exact values can be derived.     

Slow convergence of the Gibbs sampler

We consider the Gibbs sampler as a tool for generating an absolutely continuous probability measure ≥ on R^d. When an appropriate irreducibility condition is satisfied, the Gibbs Markov chain (X_n;n ≥ 0) converges in total variation to its target distribution ≥. Sufficient conditions for geometric convergence have been given by various authors. Here we illustrate, by means of simple examples, how slow the convergence can be. In particular, we show that given a sequence of positive numbers decreasing to zero, say (b_n;n ≥ 1), one can construct an absolutely continuous probability measure ≥ on R^d which is such that the total variation distance between ≥ and the distribution of X_n, converges to 0 at a rate slower than that of the sequence (b_n;n ≥ 1). This can even be done in such a way that ≥ is the uniform distribution over a bounded connected open subset of R^d. Our results extend to hit-and-run samplers with direction distributions having supports with symmetric gaps.     

A Bayesian analysis for the Wilcoxon signed-rank statistic

A Bayesian analysis is provided for the Wilcoxon signed-rank statistic (T⁺). The Bayesian analysis is based on a sign-bias parameter φ on the (0, 1) interval. For the case of a uniform prior probability distribution for φ and for small sample sizes (i.e., 6 ? n ? 25), values for the statistic T⁺ are computed that enable probabilistic statements about φ. For larger sample sizes, approximations are provided for the asymptotic likelihood function P(T⁺|φ) as well as for the posterior distribution P(φ|T⁺). Power analyses are examined both for properly specified Gaussian sampling and for misspecified non Gaussian models. The new Bayesian metric has high power efficiency in the range of 0.9–1 relative to a standard t test when there is Gaussian sampling. But if the sampling is from an unknown and misspecified distribution, then the new statistic still has high power; in some cases, the power can be higher than the t test (especially for probability mixtures and heavy-tailed distributions). The new Bayesian analysis is thus a useful and robust method for applications where the usual parametric assumptions are questionable. These properties further enable a way to do a generic Bayesian analysis for many non Gaussian distributions that currently lack a formal Bayesian model.     

Selecting the normal population with the smallest variance: A restricted subset selection rule

Consider k( ? 2) normal populations whose means are all known or unknown and whose variances are unknown. Let σ²_[1] ? ??? ? σ_[k]² denote the ordered variances. Our goal is to select a non empty subset of the k populations whose size is at most m(1 ? m ? k ? 1) so that the population associated with the smallest variance (called the best population) is included in the selected subset with a guaranteed minimum probability P* whenever σ²_[2]/σ_[1]² ? δ* > 1, where P* and δ* are specified in advance of the experiment. Based on samples of size n from each of the populations, we propose and investigate a procedure called R_BCP. We also derive some asymptotic results for our procedure. Some comparisons with an earlier available procedure are presented in terms of the average subset sizes for selected slippage configurations based on simulations. The results are illustrated by an example.     

On tests of symmetry,marginal homogeneity and quasi-symmetry in two-way contingency tables based on minimum φ-divergence estimator with constraints

The restricted minimum φ-divergence estimator, [Pardo, J.A., Pardo, L. and Zografos, K., 2002, Minimum φ-divergence estimators with constraints in multinomial populations. Journal of Statistical Planning and Inference, 104, 221–237], is employed to obtain estimates of the cell frequencies of an I×I contingency table under hypotheses of symmetry, marginal homogeneity or quasi-symmetry. The associated φ-divergence statistics are distributed asymptotically as chi-squared distributions under the null hypothesis. The new estimators and test statistics contain, as particular cases, the classical estimators and test statistics previously presented in the literature for the cited problems. A simulation study is presented, for the symmetry problem, to choose the best function φ₂ for estimation and the best function φ₁ for testing.     

News about Members

In this article a natural extension of the beta-binomial distribution is developed. Forced binary choice situations are modeled such that each individual has a probability p of knowing the correct answer. (This probability is distributed f(p) across the population.) Hence each individual will guess at the correct answer with probability 1 – p. The observable random variable R, the total number of correct answers (both by knowing and guessing) out of k trials has a rather complicated distribution. However, when f(p) is distributed beta with parameters m and n, the distribution P(r; k, m, n) can be expressed in terms of the well-known Gaussian hypergeometric function. This distribution has implications for true-false tests, taste tests, and virtually every other forced binary choice situation.     

Exploring the distribution for the estimator of Rosenthal's ‘fail-safe’ number of unpublished studies in meta-analysis

The present article discusses the statistical distribution for the estimator of Rosenthal's ‘file-drawer’ number N_R, which is an estimator of unpublished studies in meta-analysis. We calculate the probability distribution function of N_R. This is achieved based on the central limit theorem and the proposition that certain components of the estimator N_R follow a half-normal distribution, derived from the standard normal distribution. Our proposed distributions are supported by simulations and investigation of convergence.