• Communications for Statistical Applications and Methods
• /
• 제8권1호
• /
• pp.137-145
• /
• 2001

There are many situations in which the well-known tests such as log-rank test and Gehan-Wilcoxon test fail to detect the survival differences. Assuming large samples, these tests are developed asymptotically normal properties. Thus, they shall be called asymptotic tests in this paper, Several asymptotic tests sensitive to some specific types of survival differences have been recently proposed. This paper compares by simulations the test levels and the powers of the conventional asymptotic tests and their random permutation versions. Simulation studies show that the random permutation tests possess competitive powers compared to the corresponding asymptotic tests, keeping exact test levels even in the small sample case. It also provides the guidelines for choosing the valid and most powerful test under the given situation.

• Communications for Statistical Applications and Methods
• /
• 제18권6호
• /
• pp.837-849
• /
• 2011

본 연구에서는 자기조직화 신경망이 필요한 노드만을 가지고 최적화하여 정규혼합분포를 추정하는 모형을 제안한다. 제안한 모형은 SOMN모형과 벌점가능도를 사용한 모형을 결합한 것이다. SOMN의 장점은 수렴속도가 빠르고 표본의 크기가 작아도 발산하는 가능성이 낮다는 것이며, 벌점가능도를 사용한 모형은 필요없는 성분의 수를 줄일 수 있다는 것이다. 모의실험을 통하여 제안한 모형이 기대한 결과를 얻음을 확인하였다.

• 응용통계연구
• /
• 제15권1호
• /
• pp.187-199
• /
• 2002

표본의 크기가 작아 검정 통계량의 근사분포의 정확성이 의심스러울 때, 정확검정이 종종 사용된다. 정확검정의 장점은 1종의 오류 확률이 항상 유의수준보다 작거나 같음을 보장해 준다는 것이다. 본 논문에서는 정확검 정을 만드는 여러방법, 계산 알고리듬, 그리고 상업용 소프트웨어를 살펴보겠다. 그리고 정확검정에서 얻어지는 exact p-value와 원래 우도(true likelihood)에서 얻어지는 true p-value와의 관계도 살펴보겠다.

• 한국통신학회논문지
• /
• 제36권7C호
• /
• pp.428-434
• /
• 2011

Recently, subspace analysis has raised its performance to a higher level through the adoption of kernel-based nonlinearity. Especially, the radial basis function, based on its nonparametric nature, has shown promising results in face recognition. However, due to the endemic small sample size problem of face data, the conventional kernel-based feature extraction methods have difficulty in data representation. In this paper, we introduce a novel variant of the RBF kernel to alleviate this problem. By adopting the concept of the nearest feature line classifier, we show both effectiveness and generalizability of the proposed method, particularly regarding the small sample size issue.

• Journal of Magnetics
• /
• 제3권4호
• /
• pp.127-133
• /
• 1998

I have overviewed the change in GMR on annealing, in conjunction with the change in microstructure. The Co46Al19O35 granular thin films were annealed at 30$0^{\circ}C$ for various annealing time to change the microstructure. The magnitude of GMR decreases considerably with increasing annealing time, although the size of Co granules estimated from TEM observation show a small change. Parameter fits of magnetization curves and magnetoresistance curves to the Langevin function suggest that large clusters consisting of several small Co granules, which are coupled ferromagnetically, are related with the decrease of GMR on annealing. The temperature dependence of electrical resistivity ($\rho$) shows the relationship of log $\rho$ versus $T^{-1/2}$ for the sample annealed for 10 min., 1 hr. and 6 hrs. However, the sample annealed for 38 hrs. shows the relationship of log $\rho$ versus$ T^{-1/4},$ which represents a significant change in the transport mechanism.

• 한국신뢰성학회지:신뢰성응용연구
• /
• 제13권2호
• /
• pp.99-107
• /
• 2013

One-shot device is required to successfully perform its function only once at the moment of use. The reliability of a one-shot device should be expressed as a probability of success. In this paper, we propose a bayesian approach for estimating reliability of one-shot devices with small sample size. We employ a gamma prior to obtain the posterior distribution. Finally, we compare the accuracy of the proposed method with general maximum likelihood method.

• 대한안전경영과학회:학술대회논문집
• /
• 대한안전경영과학회 1999년도 추계학술대회
• /
• pp.247-257
• /
• 1999

There are several models for process quality assurance by quality system(ISO 9000), process capability analysis, acceptance control chart and so on. When a high level process capability has been achieved, it takes a long time to monitor the process shift, so it is sometimes necessary to develop a quicker monitoring system. To achieve a quicker quality assurance model for high-reliability process, this paper presents a model for process quality assurance when the fraction nonconforming is very small. We design an acceptance control chart based on variable quality characteristic and time-censored accelerated testing. The distribution of the characteristics is assumed to be normal of lognormal with a location parameter of the distribution that is a linear function of a stress. The design parameters are sample size, control limits and sample proportions allocated to low stress. These parameters are obtained under minimization of the relative variance of the MLE of location parameter subject to APL and RPL constraints.

• Journal of the Korean Statistical Society
• /
• 제32권1호
• /
• pp.33-46
• /
• 2003

The continual reassessment method (CRM) provides a Bayesian estimation of the maximum tolerated dose (MTD) in phase I clinical trials. The CRM has been proposed as an alternative design of the standard design. The CRM has been modified to improve practical feasibility and, recently, the likelihood approach CRM has been proposed. In this paper we investigate the consistency and asymptotic normality of the modified likelihood approach CRM in which the maximum likelihood estimate is used instead of the posterior mean. Small-sample properties of the consistency is examined using complete enumeration. Both the asymptotic results and their small-sample properties show that the modified CRML outperforms the standard design.

• Communications for Statistical Applications and Methods
• /
• 제23권6호
• /
• pp.479-496
• /
• 2016

In this survey, we present a modified inverse moment estimation of parameters and its applications. We use a specific model to demonstrate its principle and how to apply this method in practice. The estimation of unknown parameters is considered. A necessary and sufficient condition for the existence and uniqueness of maximum-likelihood estimates of the parameters is obtained for the classical maximum likelihood estimation. Inverse moment and modified inverse moment estimators are proposed and their properties are studied. Monte Carlo simulations are conducted to compare the performances of these estimators. As far as the biases and mean squared errors are concerned, modified inverse moment estimator works the best in all cases considered for estimating the unknown parameters. Its performance is followed by inverse moment estimator and maximum likelihood estimator, especially for small sample sizes.

• 품질경영학회지
• /
• 제14권2호
• /
• pp.2-8
• /
• 1986

This thesis is concerned with the design of control chart based on the sample median which is easy to use in practical situations and to analyze the properties for non-normally distributed Weibull process. In this cases are use to the quality characteristics of the process are not normally distributed but skewed due to the intermitted production, small lot size and sample size is small one n=3 or n=5, etc. And when it relates unsymmetrically distributed process, model designed median control chart is more effective than Shewhart $\bar{x}$-chart which assumed on normal distribution, when we exactly should be known Weibull distribution or estimated. The median control chart in this thesis is more robustness compared with other conventionally developed control chart.