• Communications for Statistical Applications and Methods
• /
• v.19 no.5
• /
• pp.655-662
• /
• 2012

Classification is an important research field in pattern recognition with high-dimensional predictors. The support vector machine(SVM) is a penalized feature selector and classifier. It is based on the hinge loss function, the non-convex penalty function, and the smoothly clipped absolute deviation(SCAD) suggested by Fan and Li (2001). We developed the algorithm for the multiclass SVM with the SCAD penalty function using the local quadratic approximation. For multiclass problems we compared the performance of the SVM with the $L_1$, $L_2$ penalty functions and the developed method.

• Communications for Statistical Applications and Methods
• /
• v.16 no.5
• /
• pp.813-820
• /
• 2009

In this paper, we review the utility-based generalization of the Shannon entropy and Kullback-Leibler information measure as the U-entropy and the U-relative entropy that was introduced by Friedman et al. (2007). Then, we derive some relations between the U-relative entropy and other information measures based on a parametric family of utility functions.

• Journal of the Korean Statistical Society
• /
• v.12 no.2
• /
• pp.69-80
• /
• 1983

The purpose of this paper is to obtain representation of the mathematical special functions and the numerical values of the mean square errors for the quasi-ranges in random small smaples ($n \leq 30$) from the Weibull distribution with a shape and a scale parameters, and to estimate the scale parameter by use of unbiased estimator based on the quasi-range. It will be shown that the jackknife estimator of the range is worse than the range of random samples from the given distribution in the sense of the mean square error.

• Journal of the Korean Statistical Society
• /
• v.21 no.1
• /
• pp.70-79
• /
• 1992

The generalized linear models with random regrssors case are studied for bootstrapping. Only the natural link functions are considered. It is shown that the bootstrap approximation to the distribution of the maximum likelihood estimators is valid for almost all sample sequences. A slight extension of this model is also considered.

• Communications for Statistical Applications and Methods
• /
• v.8 no.1
• /
• pp.291-297
• /
• 2001

For a stochastic regression model in which the errors are assumed to form a stationary linear process, we show that the difference between the empirical distribution functions of the errors and the estimates of those errors converges uniformly in probability to zero at the rate of $o_{p}$ ( $n^{-}$$\frac{1}{2}$) as the sample size n increases.

• Communications for Statistical Applications and Methods
• /
• v.2 no.2
• /
• pp.52-62
• /
• 1995

Consider an estimation problem under squared error loss in a family of non-regular densities with both terminals of the support being decreasing functions of an unknown parameter. Using Karlin's(1958) technique, sufficient conditions are given for generalized Bayes estimators to be admissible for estimating an arbitrarily positive, monotone parametric function and then treat some examples which illustrate our results.

• Communications for Statistical Applications and Methods
• /
• v.9 no.2
• /
• pp.325-335
• /
• 2002

In this paper we introduce some adaptive M-estimators using selector statistics to estimate the center of symmetric and continuous underlying distributions. This selector statistics is based on the idea of Hogg(1983) and Hogg et. al. (1988) who used averages of some order statistics to discriminate underlying distributions. In this paper, we use the functions of sample quantiles as selector statistics and determine the suitable quantile points based on maximizing the distance index to discriminate distributions under consideration. In Monte Carlo study, this robust estimation method works pretty good in wide range of underlying distributions.

• Communications for Statistical Applications and Methods
• /
• v.7 no.3
• /
• pp.767-772
• /
• 2000

Zellner(1994) introduced the notion of a balanced loss function in the context of a general liner model to reflect both goodness of fit and precision of estimation. We study the perspective of unifying a variety of results both frequentist and Bayesian from Poisson distributions. We show that frequentist and Bayesian results for balanced loss follow from and also imply related results for quadratic loss functions reflecting only precision of estimation. Several examples are given for Poisson distribution.

• Communications for Statistical Applications and Methods
• /
• v.13 no.1
• /
• pp.89-99
• /
• 2006

This article deals with the problem of estimating parameters of Burr Type XII distribution, on the basis of a general progressive Type II censored sample using Bayesian viewpoints. The maximum likelihood estimator does not admit closed form but explicit sharp lower and upper bounds are provided. Assuming squared error loss and linex loss functions, Bayes estimators of the parameter k, the reliability function, and the failure rate function are obtained in closed form. Finally, a simulation study is also included.

• Journal of the Korean Statistical Society
• /
• v.26 no.4
• /
• pp.531-541
• /
• 1997

The asymptotic distribution of residual autocorrelation functions from a generalized p-order random coefficient autoregressive process (GRCA(p)) is derived. To this end, we first describe the GRCA(p) models and then consider the normalised residuals after fitting the model. This result can be applied to the residual analysis for the diagonostic purpose.