• Communications for Statistical Applications and Methods
• /
• 제15권2호
• /
• pp.213-228
• /
• 2008

Using a linear loss function, this paper considers the one-sided testing problem for the exponential distribution via the empirical Bayes(EB) approach. Based on right censored data, we propose an EB test for the exponential parameter and obtain its convergence rate and asymptotic optimality, firstly, under the condition that the censoring distribution is known and secondly, that it is unknown.

• Communications for Statistical Applications and Methods
• /
• 제24권2호
• /
• pp.129-142
• /
• 2017

In this paper we consider the problem of the model selection/discrimination among three different positively skewed lifetime distributions. Lindley, Weibull, and gamma distributions have been used to effectively analyze positively skewed lifetime data. This paper assesses how much closer the Lindley distribution gets to Weibull and gamma distributions. We consider three techniques that involve the likelihood ratio test, asymptotic likelihood ratio test, and minimum Kolmogorov distance as optimality criteria to diagnose the appropriate fitting model among the three distributions for a given data set. Monte Carlo simulation study is performed for computing the probability of correct selection based on the considered optimality criteria among these families of distributions for various choices of sample sizes and shape parameters. It is observed that overall, the Lindley distribution is closer to Weibull distribution in the sense of likelihood ratio and Kolmogorov criteria. A real data set is presented and analyzed for illustrative purposes.

• Communications for Statistical Applications and Methods
• /
• 제26권6호
• /
• pp.527-538
• /
• 2019

This paper considers the issue of obtaining the optimal design in Poisson regression model when the sample size is small. Poisson regression model is widely used for the analysis of count data. Asymptotic theory provides the basis for making inference on the parameters in this model. However, for small size experiments, asymptotic approximations, such as unbiasedness, may not be valid. Therefore, first, we employ the second order expansion of the bias of the maximum likelihood estimator (MLE) and derive the mean square error (MSE) of MLE to measure the quality of an estimator. We then define DM-optimality criterion, which is based on a function of the MSE. This criterion is applied to obtain locally optimal designs for small size experiments. The effect of sample size on the obtained designs are shown. We also obtain locally DM-optimal designs for some special cases of the model.

• Journal of the Korean Statistical Society
• /
• 제24권2호
• /
• pp.303-312
• /
• 1995

We consider the following type of general semi-parametric non-linear regression model : $y_i = f_i(\theta) + \epsilon_i, i=1, \cdots, n$ where ${f_i(\cdot)}$ represents the set of non-linear functions of the unknown parameter vector $\theta' = (\theta_1, \cdots, \theta_p)$ and ${\epsilon_i}$ represents the set of measurement errors with unknown distribution. Under suitable finite-sample, small-dispersion asymptotic framework, we derive a general lower bound for the asymptotic mean squared error (AMSE) matrix of the Gauss-consistent estimator of $\theta$. We then prove the fundamental result that the general non-linear least squares estimator (NLSE) is an optimal estimator within the class of all regular Gauss-consistent estimators irrespective of the type of the distribution of the measurement errors.

• Industrial Engineering and Management Systems
• /
• 제7권2호
• /
• pp.126-132
• /
• 2008

The bin packing problem (BPP) is an NP-Complete Problem. The problem can be described as there are $N=\{1,2,{\cdots},n\}$ which is a set of item indices and $L=\{s1,s2,{\cdots},sn\}$ be a set of item sizes sj, where $0<sj{\leq}1$, ${\forall}j{\in}N$. The objective is to minimize the number of bins used for packing items in N into a bin such that the total size of items in a bin does not exceed the bin capacity. Assume that the bins have capacity equal to one. In the past, many researchers put on effort to find the heuristic algorithms instead of solving the problem to optimality. Then, the quality of solution may be measured by the asymptotic worst-case ratio or the average-case ratio. The First Fit Decreasing (FFD) is one of the algorithms that its asymptotic worst-case ratio equals to 11/9. Many researchers prove the asymptotic worst-case ratio by using the weighting function and the proof is in a lengthy format. In this study, we found an easier way to prove that the asymptotic worst-case ratio of the First Fit Decreasing (FFD) is not more than 11/9. The proof comes from two ideas which are the occupied space in a bin is more than the size of the item and the occupied space in the optimal solution is less than occupied space in the FFD solution. The occupied space is later called the weighting function. The objective is to determine the maximum occupied space of the heuristics by using integer programming. The maximum value is the key to the asymptotic worst-case ratio.

• Journal of the Korean Statistical Society
• /
• 제28권4호
• /
• pp.415-433
• /
• 1999

This paper considers the design of step-stress accelerated life tests for the Weibull distribution with a nonconstant shape parameter under Type I censoring. It is assumed that scale and shape parameters are log-linear functions of (possibly transformed) stress and that a cumulative exposure model holds for the effect of changing stress. The asymptotic variance of the maximum likelihood estimator of a stated quantile at design stress is used as an optimality criterion. The optimum three step-stress plans are presented for selected values of design parameters and the effects of errors in pre- estimates of the design parameters are investigated.

• 대한산업공학회지
• /
• 제24권3호
• /
• pp.367-379
• /
• 1998

This paper considers two modes of partially accelerated life tests for items having Weibull lifetime distributions. In a use-to-acclerated mode each item is first run at use condition and, if it does not fail for a specified time, then it is run at accelerated condition until a predetermined censoring time. In an accelerated-to-use mode each one is first run at accelerated condition and, if it does not fail for a specified time, then it is run at use condition. Maximum likelihood estimators of the parameters of the lifetime distribution at use condition, and the 'acceleration factor' are obtained. The stress change time for each mode is determined to minimize the asymptotic variance of the acceleration factor, and the two modes are compared. For selected values of the design parameters the optimum plans are obtained, and the effects of the incorrect pre-estimates of the design parameters are investigated. Minimizing the generalized asymptotic variance of the estimators of the model parameters is also considered as an optimality criterion.

• Communications for Statistical Applications and Methods
• /
• 제3권3호
• /
• pp.225-233
• /
• 1996

In this paper, we propose Bayes estimators of the finite population mean based on heavy-tailed prior distributions using scale mixtures of normals. Also, the asymptotic optimality property of the proposed Bayes estimators is proved. A numerical example is provided to illustrate the results.

• Communications for Statistical Applications and Methods
• /
• 제8권1호
• /
• pp.87-96
• /
• 2001

We consider the empirical Bayes estimation problems with the binomial and normal components when the prior distributions are unknown but are assumed to be in certain families. There may be the families of all distributions on the parameter space or subfamilies such as the parametric families of conjugate priors. We treat both cases and establish the asymptotic optimality for the corresponding decision procedures.

• Journal of the Korean Statistical Society
• /
• 제27권3호
• /
• pp.251-264
• /
• 1998

When the LRT is not feasible, we define quasi-LRT(QLRT) as a modification of the LRT Under some appropriate conditions the QLRT shares Bahadur optimality and Bartlett Adjustability with the LRT. When we can find maximum likelihood estimator under the null parameter space but not under the unrestricted parameter space, our QLRT is Bahadur optimal as is the LRT We suggest the stopping rule of the Newton-Raphson iterations for constructing the QLRT statistics which are Bartlett adjustable.