• Journal of the Korean Data and Information Science Society
• /
• v.21 no.1
• /
• pp.163-172
• /
• 2010

In this paper, we deal with the problem for testing homogeneity of coecients of variation in several normal distributions. We propose Bayesian hypothesis testing procedures based on the Bayes factor under noninformative prior. The noninformative prior is usually improper which yields a calibration problem that makes the Bayes factor to be dened up to a multiplicative constant. So we propose the objective Bayesian hypothesis testing procedures based on the fractional Bayes factor and the intrinsic Bayes factor under the reference prior. Simulation study and a real data example are provided.

• Communications for Statistical Applications and Methods
• /
• v.10 no.3
• /
• pp.909-917
• /
• 2003

Although many classification models have been used to classify binary data, none of the classification models dominates all varying circumstances depending on the number of variables and the size of data(Asparoukhov and Krzanowski (2001)). This paper proposes a classification model which uses information on marginal distributions of sub-variables and its maximum entropy distribution. Classification experiments by using simulation are discussed.

• International Journal of Reliability and Applications
• /
• v.8 no.1
• /
• pp.17-25
• /
• 2007

Maximum likelihood method is utilized to estimate the two parameters of generalized exponential distribution based on grouped and censored data. This method does not give closed form for the estimates, thus numerical procedure is used. Reliability measures for the generalized exponential distribution are calculated. Testing the goodness of fit for the exponential distribution against the generalized exponential distribution is discussed. Relevant reliability measures of the generalized exponential distributions are also evaluated. A set of real data is employed to illustrate the results given in this paper.

• Journal of the Korean Data and Information Science Society
• /
• v.17 no.1
• /
• pp.31-39
• /
• 2006

In many cases, we frequently get a desired information based on the appropriate statistical analysis of collected data sets. Lots of statistical theory rely on the assumption of the normality of the data. In this paper, we compare the empirical power of some normality tests including sample entropy quantity. Monte carlo simulation is conducted for the calculation of empirical power of considered normality tests by varying sample sizes for various distributions.

• Journal of the Korean Data and Information Science Society
• /
• v.19 no.4
• /
• pp.1379-1390
• /
• 2008

This article deals with the problem of testing the difference of quantiles in exponential distributions. We propose Bayesian hypothesis testing procedures for the difference of two quantiles under the noninformative prior. The noninformative prior is usually improper which yields a calibration problem that makes the Bayes factor to be defined up to a multiplicative constant. So we propose the objective Bayesian hypothesis testing procedures based on the fractional Bayes factor and the intrinsic Bayes factor under the matching prior. Simulation study and a real data example are provided.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.631-634
• /
• 2004

In this paper, the Bayesian recurrent neural network (BRNN) is proposed to predict time series data. Among the various traditional prediction methodologies, a neural network method is considered to be more effective in case of non-linear and non-stationary time series data. A neural network predictor requests proper learning strategy to adjust the network weights, and one need to prepare for non-linear and non-stationary evolution of network weights. The Bayesian neural network in this paper estimates not the single set of weights but the probability distributions of weights. In other words, we sets the weight vector as a state vector of state space method, and estimates its probability distributions in accordance with the Bayesian inference. This approach makes it possible to obtain more exact estimation of the weights. Moreover, in the aspect of network architecture, it is known that the recurrent feedback structure is superior to the feedforward structure for the problem of time series prediction. Therefore, the recurrent network with Bayesian inference, what we call BRNN, is expected to show higher performance than the normal neural network. To verify the performance of the proposed method, the time series data are numerically generated and a neural network predictor is applied on it. As a result, BRNN is proved to show better prediction result than common feedforward Bayesian neural network.

• Journal of Welding and Joining
• /
• v.9 no.1
• /
• pp.40-47
• /
• 1991

The most fatigue tests carried out under the either stress or strain control, but machines and structures had taken variable stress. This variable stress was treated as statistics based on p-type distributions. In this paper, the cumulative fatigue damage of SM45C round bar specimens having a center hole resulting from block loading with p-distributions in rotating bending conditions, is presented. The value of p was changed in the range from 0.25 to 1; 0.25, 0.5, 0.75, 1. The following conclusions were obtained through the constant stress amplitude experiments and the block loading experiments. (1) In constant loading test, fatigue life was affected by cyclic rate. From experimental data, N$_{f}$ (100cpm)/N$_{f}$(3000cpm)equal to 0.56. (2) In case of the cyclic rate 100cpm and 3000cpm, at the high stress amplitude level the crack propagation life N$_{*}$f is longer than the low stress amplitude level. (3) Miner's hypothesis may be valid for p=0.75 and prediction of fatigue life by Haibach's method agree with experimental data well for the case p=0.5, while the modified Miner's method agree with experimental data well for the case p=0.25.5.

• Journal of Mechanical Science and Technology
• /
• v.20 no.3
• /
• pp.335-344
• /
• 2006

A comprehensive review of through thickness transverse residual stress distributions in a range of as-welded and mechanically bent components made up of a range of steels has been carried out, and simplified generic transverse residual stress profiles for a plate and pipe components have been proposed. The geometries consisted of welded pipe butt joints, T-plate joints, tubular T-joints, tubular Y-joints and a pipe on plate joints as well as cold bent tubes and pipes. The collected data covered a range of engineering steels including ferritic, austenitic, C-Mn and Cr-Mo steels. Measured residual stress data, normalised with respect to the parent material yield stress, has shown a good linear correlation versus the normalised depth of the region containing the residual stress resulting from the welding or cold-bending process. The proposed simplified generic residual stress profiles based on the mean statistical linear fit of all the data provides a reasonably conservative prediction of the stress intensity factors. Whereas the profiles for the assessment procedures are fixed and case specific, the simple bilinear profiles for the residual stresses obtained by shifting the mean and bending stress from the mean regression line have been proposed and validated.

• The Korean Journal of Applied Statistics
• /
• v.25 no.5
• /
• pp.757-773
• /
• 2012

Traditional value at risk(S-VaR) has a difficulity in predicting the future risk of financial asset prices since S-VaR is a backward looking measure based on the historical data of the underlying asset prices. In order to resolve the deficiency of S-VaR, an economic value at risk(E-VaR) using the risk neutral probability distributions is suggested since E-VaR is a forward looking measure based on the option price data. In this study E-VaR is estimated by assuming the generalized gamma distribution(GGD) as risk neutral density function which is implied in the option. The estimated E-VaR with GGD was compared with E-VaR estimates under the Black-Scholes model, two-lognormal mixture distribution, generalized extreme value distribution and S-VaR estimates under the normal distribution and GARCH(1, 1) model, respectively. The option market data of the KOSPI 200 index are used in order to compare the performances of the above VaR estimates. The results of the empirical analysis show that GGD seems to have a tendency to estimate VaR conservatively; however, GGD is superior to other models in the overall sense.

• Geophysics and Geophysical Exploration
• /
• v.16 no.1
• /
• pp.1-5
• /
• 2013

Stacking in seismic processing plays an important role in improving signal-to-noise ratio and imaging quality of seismic data. However, the conventional stacking method doesn't remove random noises with various distributions and outliers up to a satisfactory level. This paper introduces a robust and optimum weighted stack method which shows both robustness to outlier noises and optimum in removing random noises. This was achieved by combining the robust median stacking with the optimum weighted stacking using local correlation. Application of the method to synthetic data showed that the proposed method is very effective in suppressing random noises with various distributions including outliers.