• Speech Sciences
• /
• v.9 no.2
• /
• pp.135-146
• /
• 2002

Generally a speaker verification system improves its system recognition ratio by regularizing log likelihood ratio, using a speaker model and its background speaker model that are required to be verified. The speaker-based cohort method is one of the methods that are widely used for selecting background speaker model. Recently, Gaussian-based cohort model has been suggested as a virtually synthesized cohort model, and unlike a speaker-based model, this is the method that chooses only the probability distributions close to basic speaker's probability distribution among the several neighboring speakers' probability distributions and thereby synthesizes a new virtual speaker model. It shows more excellent results than the existing speaker-based method. This study compared the existing speaker-based background speaker models and virtual speaker models and then constructed new virtual background speaker model groups which combined them in a certain ratio. For this, this study constructed a speaker verification system that uses GMM (Gaussin Mixture Model), and found that the suggested method of selecting virtual background speaker model shows more improved performance.

• Communications for Statistical Applications and Methods
• /
• v.15 no.3
• /
• pp.429-440
• /
• 2008

In this paper, we develop the noninformative priors when the parameter of interest is the common coefficient of variation in two inverse Gaussian distributions. We want to develop the first and second order probability matching priors. But we prove that the second order probability matching prior does not exist. It turns out that the one-at-a-time and two group reference priors satisfy the first order matching criterion but Jeffreys' prior does not. The Bayesian credible intervals based on the one-at-a-time reference prior meet the frequentist target coverage probabilities much better than that of Jeffreys' prior. Some simulations are given.

• Journal of the Korean Statistical Society
• /
• v.33 no.1
• /
• pp.107-128
• /
• 2004

Bivariate Laplace distributions for which both marginal distributions and Laplace are discussed. Three kinds of bivariate Laplace distributions which are extended bivariate exponential distributions of Gumbel (1960) are introduced in this paper. These symmetrical distributions are compared with asymmetrical distributions of Kotz et al. (2000). Their probability density functions, cumulative distribution functions are derived. Conditional skewnesses and kurtoses are also defined. Their correlation coefficients are calculated and compared with others. We proposed bivariate random vector generating methods whose distributions are bivariate Laplace. With sample means and medians obtained from generated random vectors, variance and covariance matrices of means and medians are calculated and discussed with those of bivariate normal distribution.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.10a
• /
• pp.15-19
• /
• 2003

In this paper, we develop the Jeffreys' prior, reference priors and the probability matching priors for the intraclass correlation coefficient of a symmetric normal distribution. We next verify propriety of posterior distributions under those noninformative priors. We examine whether reference priors satisfy the probability matching criterion.

• Water for future
• /
• v.28 no.5
• /
• pp.191-203
• /
• 1995

In this study, the past flood levels of Goan station, which is one of major gaging stations and located at downstream of Paldang dam, were converted based on the 1994's cross section and the flood quantiles were estimated from flood frequency analysis. The recently established rating curve was used to convert flood levels. And the parameters of the several probability distributions commonly used in hydrologic analysis were estimated based on the method of probability weighted moments and the goodness of fit tests were applied to those distributions. As a result, the gamma-2 and gamma-3 distributions were selected as the appropriate models. The flood lovels and quantiles for selected return periods were calculated based on those distributions. Furthermore, frequency analysis using historical flood information was performed to overcome the misleading caused by missing data.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.7
• /
• pp.175-182
• /
• 2020

Fisher's classical method for combining independent p-values from continuous distributions is widely used but it is known to be inadequate for combining p-values from discrete probability distributions. Instead, the discrete analog of Fisher's classical method is used as an alternative for combining p-values from discrete distributions. In this paper, firstly we obtain p-values from discrete probability distributions associated with multi-sample location test data (Fisher-Pitman test and Kruskall-Wallis test data) by permutation method, and secondly combine the permutaion p-values by the discrete analog of Fisher's classical method. And we finally compare the combined p-values from both the discrete analog of Fisher's classical method and Fisher's classical method.

• Nuclear Engineering and Technology
• /
• v.54 no.5
• /
• pp.1606-1615
• /
• 2022

For investigating whether the MARSSIM nonparametric test has sufficient statistical power when a site has a specific contamination distribution before conducting a final status survey (FSS), a novel approach was proposed to predict the release probability of the site. Five distributions were assumed: lognormal distribution, normal distribution, maximum extreme value distribution, minimum extreme value distribution, and uniform distribution. Hypothetical radioactivity populations were generated for each distribution, and Sign tests were performed to predict the release probabilities after extracting samples using Monte Carlo simulations. The designed Type I error (0.01, 0.05, and 0.1) was always satisfied for all distributions, while the designed Type II error (0.01, 0.05, and 0.1) was not always met for the uniform, maximum extreme value, and lognormal distributions. Through detailed analyses for lognormal and normal distributions which are often found for contaminants in actual environmental or soil samples, it was found that a greater statistical power was obtained from survey units with normal distribution than with lognormal distribution. This study is expected to contribute to achieving the designed decision error when the contamination distribution of a survey unit is identified, by predicting whether the survey unit passes the statistical test before undertaking the FSS according to MARSSIM.

• Journal of the Korean Statistical Society
• /
• v.29 no.3
• /
• pp.315-318
• /
• 2000

In this note we use the Poisson Summation Formula to generalize a result of Harris and Park (1994) on lattice distributions induced by uniform (0,1) random variables to those generated by random variables with step functions as their probability functions.

• Journal of the Korean Statistical Society
• /
• v.28 no.4
• /
• pp.435-441
• /
• 1999

A 'convergence in probability' rate of the empirical distributions and quantiles of linear processes is obtained. As an application of the limit theorems, a trimmed mean for the location of the linear process is considered. It is shown that the trimmed mean is asymptotically normal. A consistent estimator for the asymptotic variance of the trimmed mean is provided.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.05a
• /
• pp.109-114
• /
• 2003

The concept of dispersion is intrinsic to the theory and practice of statistics. A formulation of the concept of dispersion can be obtained by comparing the probability of intervals centered about a location parameter, which is peakedness ordering introduced first by Birnbaum (1948). We consider statistical inference concerning peakedness ordering between two arbitrary distributions. We propose nonparametric maximum likelihood estimator of two distributions under peakedness ordering and a likelihood ratio test for equality of dispersion in the sense of peakedness ordering.