• Structural Engineering and Mechanics
• /
• v.28 no.3
• /
• pp.263-280
• /
• 2008

Information on the distribution of the basic random variable is essential for the accurate analysis of structural reliability. The usual method for determining the distributions is to fit a candidate distribution to the histogram of available statistical data of the variable and perform approximate goodness-of-fit tests. Generally, such candidate distribution would have parameters that may be evaluated from the statistical moments of the statistical data. In the present paper, a cubic normal distribution, whose parameters are determined using the first four moments of available sample data, is investigated. A parameter table based on the first four moments, which simplifies parameter estimation, is given. The simplicity, generality, flexibility and advantages of this distribution in statistical data analysis and its significance in structural reliability evaluation are discussed. Numerical examples are presented to demonstrate these advantages.

• The Korean Journal of Applied Statistics
• /
• v.4 no.1
• /
• pp.33-45
• /
• 1991

In the analysis of changepoint model the situation where single observation is taken at each time point has been considered. In an effort to extend this to the general situation, we may consider the changepoint model with more than one observation at each time point. These tests are developed without assuming any particular form for the underlying distribution, we propose the one-sided and two-sided nonparametric tests by extending the tests that have been considered in the changepoint model with single observation at each time point and obtain their asymptotic null distributions. We compare the empirical powers among the extended changepoint tests under one-sided or two-sided alternatives. We also compare the powers of the extended changepoint tests with those of the original test via the Monte Carlo simulation.

• The Korean Journal of Applied Statistics
• /
• v.15 no.1
• /
• pp.187-199
• /
• 2002

When the sample size is small, exact tests are often employed because the asymptotic distribution of the test statistic is in doubt. The advantage of exact tests is that it is guaranteed to bound the type I error probability to the nominal level. In this paper we review the methods of constructing exact tests, the algorithm and commercial software. We also examine the difference between exact p-values obtained from exact tests and true p-values obtained from the true underlying distribution.

• Transactions on Control, Automation and Systems Engineering
• /
• v.2 no.1
• /
• pp.40-46
• /
• 2000

In this paper, a transient identification based on a Hidden Markov Model (HMM) has been suggested and evaluated experimentally for the classification of transients in the dynamic process. The transient can be identified by its unique time dependent patterns related to the principal variables. The HMM, a double stochastic process, can be applied to transient identification which is a spatial and temporal classification problem under a statistical pattern recognition framework. The HMM is created for each transient from a set of training data by the maximum-likelihood estimation method. The transient identification is determined by calculating which model has the highest probability for the given test data. Several experimental tests have been performed with normalization methods, clustering algorithms, and a number of states in HMM. Several experimental tests have been performed including superimposing random noise, adding systematic error, and untrained transients. The proposed real-time transient identification system has many advantages, however, there are still a lot of problems that should be solved to apply to a real dynamic process. Further efforts are being made to improve the system performance and robustness to demonstrate reliability and accuracy to the required level.

• Journal of Korean Society for Quality Management
• /
• v.29 no.3
• /
• pp.49-65
• /
• 2001

This paper presents new compromise ALT plan which is applied to situations that true relationship between stress and parameters is not known exactly. The assumed failure distribution of this study is one of location-scale family, i. e., exponential, Weibull, and lognormal distributions which have been ones of the popular choices of failure distributions. The method of applying the stress is constant, and the censoring mechanism is Type I censoring. Compared with existing compromise plans under true simple linear model in terms of statistical efficiency, the efficiency of new compromise plan is better than the corresponding other compromise ones in most cases. For case when true model is quadratic, this plan can be used without any severe loss in statistical efficiency. The proposed new compromise ALT plan is illustrated with a numerical example and sensitivity analyses are conducted to study effects of pre-estimates of design parameters.

• Journal of the Korean Statistical Society
• /
• v.28 no.4
• /
• 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.

• Communications for Statistical Applications and Methods
• /
• v.3 no.3
• /
• pp.205-214
• /
• 1996

As flexible Bayesian test criteria for nested point null hypotheses of multiple regression coefficients, partial and overall Bayes factors are introduced under a class of intuitively meaningful prior. The criteria lead to a simple method for considering different prior beliefs on the subspaces that constitute a partition of the coefficient parameter space. A couple of tests are suggested based on the criteria. It is shown that they enable us to obtain pairwise comparisons of hypotheses of the partitioned subspaces. Through a Monte Carlo simulation, performance of the tests based on the criteria are compared with the usual Bayesian test (based on Bayes factor)in terms of their respective powers.

• Proceedings of the KIEE Conference
• /
• 1987.07b
• /
• pp.1394-1397
• /
• 1987

It is difficult to detect edge segments from a noisy image since the image have a noise in piratical applications which utilize some type of visual input capability. Hence, the proposed algorithm consists of the modality tests based on parallel statistical tests without a noise removal preprocessing or postprocessing, and the edge detection technique With one-Pixel edge segments in this paper. The algorithm is very reliable and effective in the case of those situations where the Picture is poor quality and low resolution. And it does'nt require thinning operation and thresholding in hand. Experimental comparision With the more conventional techniques when applied to typical low-quality Pictures confirms good capabilities of the algorithm.

• Communications for Statistical Applications and Methods
• /
• v.12 no.1
• /
• pp.207-216
• /
• 2005

A binary segmentation procedure is a simple recursive approach to detect clusters and provide inferences for the study space when the shape of the clusters and the number of clusters are unknown. The procedure involves a sequence of nested hypothesis tests of a single cluster versus a pair of distinct clusters. The size and the shape of the clusters evolve as the procedure proceeds. The procedure allows for various growth clusters and for arbitrary baseline densities which govern the form of the hypothesis tests. A real tree data is used to highlight the procedure.

• Communications for Statistical Applications and Methods
• /
• v.12 no.1
• /
• pp.101-115
• /
• 2005

We propose in this article a procedure for testing unit and fractional orders of integration, with the roots simultaneously occurring in the trend, the seasonal and the cyclical component of the time series. The tests have standard null and local limit distributions. However, finite sample critical values are computed, and several Monte Carlo experiments conducted across the paper show that the rejection frequencies against unit (and fractional) orders of integration are relatively high in all cases. The tests are applied to the UK consumption and income series, the results showing the importance of the roots corresponding to the trend and the seasonal components and, though the unit roots are found to be fairly suitable models, we show that fractional processes (including one for the cyclical component) may also be plausible alternatives in some cases.