• Journal of the Korean Data and Information Science Society
• /
• 제16권2호
• /
• pp.371-382
• /
• 2005

The paper compares the performance of some widely used Bayesian estimators such as Bayes estimator, empirical Bayes estimator, constrained Bayes estimator and constrained Bayes estimator by means of a new measurement under squared error loss function for the typical normal-normal situation. The proposed measurement is a weighted sum of the precisions of first and second moments. As a result, one can gets the criterion according to the size of prior variance against the population variance.

• Journal of the Korean Data and Information Science Society
• /
• 제10권1호
• /
• pp.11-18
• /
• 1999

To estimate the mean and the variance of a lognormal distribution, Finney (1941) derived the uniformly minimun variance unbiased estimators(UMVUE) in the form of infinite series. However, the conditions ${\sigma}^{2}\;>\;n\;and\;{\sigma}^{2}\;<\;\frac{n}{4}$ for computing $E(\hat{\theta}_{AM})\;and\;E(\hat{\eta}^{2}_{AM})$ are necessary. In this paper, we give an alternative derivation of the UMVUE's.

• 응용통계연구
• /
• 제3권2호
• /
• pp.79-89
• /
• 1990

정규분포를 가정한 통상적인 일원배치모형에서 모평균들을 비교하는 부분집합 선택방법으로서 베이즈-$P^*$ 선택방법을 제시하고 기존의 방법과의 관계를 알아보고, 그 운용특성에 대한 모의실험의 결과를 고찰하였다.

• 산업공학
• /
• 제7권1호
• /
• pp.75-80
• /
• 1994

In a recent article, Leon et al. lucidly explained the ideas of the Taguchi two-stage procedure for parameter design optimization, and proposed alternative performance measures called PerMIA to the signal-to-noise ratios. On the other hand, Box proposed an empirical approach to the problem based upon monotone transformations of the performance characteristic(y). This paper develops procedures for parameter design optimization under the assumptions that the expected loss(not necessarily a mean squared error loss) is increasing with respect to the variance of the error in y, and that the mean of y satisfies certain conditions of adjustability. It turns out that the variance of the error in y can play the role of PerMIA, and it is further shown that the derived PerMIA can be adapted to the Box empirical procedure for the minimization of the expected loss in the original metric.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2002년도 11월 학술대회지
• /
• pp.197-200
• /
• 2002

In this paper, we propose a phoneme duration modeling in a speech recognition system based on disicion tree state tying. We assume that phone duration has a Gamma distribution. In a training mode, we model mean and variance of each state duration in context-independent phone model based on decision tree state tying. In a recognition mode, we get mean and variance of each context-dependent phone duration form state duration information obtaind during training mode. We make a comparative study of the proposed meth with conventinal methods. Our method results in good performance compared with conventional methods.

• 산업경영시스템학회지
• /
• 제19권40호
• /
• pp.179-186
• /
• 1996

Previous researches on optimum ALT plans have shown how to find optimum allocation, lowest stress and sample size subject to minimizing the variance of mean life estimate. Those researches have assumed the censoring time constant. But under certain circumstances such assumptions may not be reasonable. In this paper the constraint on constant censoring time is relaxed, and censoring time is considered variable. If all other test parameters remain unvaried, the variance of the mean life estimate reciprocally increases of the censoring time. This idea in this paper extends to the optimum censoring time. Thus, this paper presents the optimal censoring time in such situations, and finds the optimal lowest stress and allocation based on that optimal censoring time.

• Journal of the Korean Data and Information Science Society
• /
• 제21권1호
• /
• pp.155-162
• /
• 2010

In this paper we propose a doubly penalized kernel method which estimates both the mean function and the variance function simultaneously by kernel machines for heteroscedastic autoregressive data. We also present the model selection method which employs the cross validation techniques for choosing the hyper-parameters which aect the performance of proposed method. Simulated examples are provided to indicate the usefulness of proposed method for the estimation of mean and variance functions.

• 대한기계학회논문집
• /
• 제7권4호
• /
• pp.451-460
• /
• 1983

Computational four-equation turbulence model is developed and is applied to predict twodimensional unsteady thermal surface discharge into a reservoir. Turbulent stresses and heat fluxes in the momentum and energy equations are determined from transport equations for the turbulent kinetic energy (R), isotropic rate of kinetic energy dissipation (.epsilon.), mean square temperature variance (theta. over bar $^{2}$), and rate of destruction of the temperature variance (.epsilon. $_{\theta}$). Computational results by four-equation model are favorably compared with those obtained by an extended two-equation model. Added advantage of the four-equation model is that it yields quantitative information about the ratio between the velocity time scale and the thermal time scale and more detailed information about turbulent structure. Predicted time scale ratio is within experimental observations by others. Although the mean velocity and temperature fields are similarly predicted by both models, it is found that the four-equation model is preferably candidate for prediction of highly buoyant turbulent flows.

• 품질경영학회지
• /
• 제20권1호
• /
• pp.1-10
• /
• 1992

When we compare different control charting schemes, the average run length of each control chart is usually used. The use of the average run length implies that there is unbounded number of samples or observations. The regression recursive residuals, however, have been applied to the cumulative sum chart to detect whether the mean or variance changes. To implement choice of decision interval, we calculate the probability that certain fixed number of control statistics stay in the in-control state. This probability can be used as the significance level of a test for detecting the change in the residual mean or variance of the data with a finite number of observations.

• 대한산업공학회지
• /
• 제17권1호
• /
• pp.95-108
• /
• 1991

For the case where the lifetime at a constant stress level has exponential distribution, optimal accelerated life test plans are developed under the assumptions of intermittent inspection and Type I censoring. In a optimal plan, the low and high stress levels, the proportion of test units allocated and the inspection times at each stress are determined such that the asymptotic variance of the maximum likelihood estimator of logarithmic transformed mean at the use condition is minimized. In addition to the optimal plan in which numerical technique to solve the set of nonlinear equations must be employed to determine inspection times at each stress level, we also propose another plans which employ equally-spaced or equal probability inspection schemes at two overstress levels of corresponding optimal one. For both optimal and proposed plans, computational results indicate that the asymptotic variance of the estimated mean at the use stress is insensitive to number of inspections at overstress levels for the range of parameter values considered.