• The Korean Journal of Applied Statistics
• /
• v.27 no.3
• /
• pp.397-406
• /
• 2014

The minimum density power divergence estimation has been a popular topic in the field of robust estimation for since Basu et al. (1988). The minimum density power divergence estimator has strong robustness properties with the little loss in asymptotic efficiency relative to the maximum likelihood estimator under model conditions. However, a limitation in applying this estimation method is the algebraic difficulty on an integral involved in an estimation function. This paper considers a minimum density power divergence estimation method with approximated divergence avoiding such difficulty. As an example, we consider the normal-exponential convolution model introduced by Bolstad (2004). The estimated divergence in this case is too complicated; consequently, a Laplace approximation is employed to obtain a manageable form. Simulations and an empirical study show that the minimum density power divergence estimators based on an approximated estimated divergence for the normal-exponential model perform adequately in terms of bias and efficiency.

• The Korean Journal of Applied Statistics
• /
• v.35 no.5
• /
• pp.631-644
• /
• 2022

This paper considers robust estimation of the sparse vector autoregressive model (sVAR) useful in high-dimensional time series analysis. First, we generalize the result of Xu et al. (2008) that the adaptive lasso indeed has robustness in sVAR as well. However, adaptive lasso method in sVAR performs poorly as the number and sizes of outliers increases. Therefore, we propose new robust estimation methods for sVAR based on least absolute deviation (LAD) and Huber estimation. Our simulation results show that our proposed methods provide more accurate estimation in turn showed better forecasting performance when outliers exist. In addition, we applied our proposed methods to power usage data and confirmed that there are unignorable outliers and robust estimation taking such outliers into account improves forecasting.

• The Korean Journal of Applied Statistics
• /
• v.29 no.3
• /
• pp.463-470
• /
• 2016

In analysing a time series on the frequency domain, the spectral estimator (or periodogram) is a very useful statistic to identify the periods of a time series. However, the spectral estimator is very sensitive in nature to outliers, so that the spectral estimator in terms of M-estimation has been studied by some researchers. Pak (2001) proposed an empirical method to choose a tuning parameter for the Huber's M-estimating function. In this article, we try to implement Pak's estimation proposal in the spectral estimator. We use the Korean housing price index as an example data set for comparing various M-estimating results.

• The Korean Journal of Applied Statistics
• /
• v.8 no.2
• /
• pp.77-86
• /
• 1995

For the problem of testing the parallelism of two regression lines, a robust procedure is proposed and examined. The proposed test statistic is based on the one-step GM-estimators of slope parameters proposed by Song et al. (1994b). These GM-estimators used the Least Trimmed Squares estimates as an initial values so as to obtain high breakdown point. Through a small-sample Monte Carlo simulation the empirical levels and powers of the proposed test are compared with other tests under various error distributions.

• The Journal of the Acoustical Society of Korea
• /
• v.11 no.1E
• /
• pp.73-78
• /
• 1992

시벼형 신호인 음성 신호의 분석에 칼만필터를 이용하였다. 일반적인 음성 분석은 프레임단위의 처리방법인 선형 예측 부호화 기법을 주로 이용하지만 음성의 시변 특성을 파악하는데에는 적절하지 못 하다. 따라서 순차적인 추정기법으로 많이 이용되는 칼만 필터를 음성 분석에 적용하였다. 또한 음성과 같은 시변신호에서는 과거 신호의 잡음의 분산값에 적당한 가중치를 부가하므로써 과거의 신호에 의해 서 현재의 추정값에 미치는 영향을 줄였으며 이를 음성의 천이 구간에서의 파라메타 추정에 사용하였 다. 그리고 음성신호 모델에서 생기는 모델링 오차는 일반적으로 백색 가우시안 잡음으로 가정하고 있 으나 이는 자음과 같은 무성음에서 특징 파라메타 푸정에는 오차가 적지만 모음등의 유성음에서는 음성 발생시의 여기신호인 펄스열에 의해서 많은 모델링 오차를 생기게 한다. 따라서 모델링 오차신호는 Non-Gaussian 확률분포로 가정한 후 로버스트 칼만 필터를 사용하여 합성으멩 대해 특징 파라메터를 추출하였다.

• Proceedings of the Korean Statistical Society Conference
• /
• 2003.05a
• /
• pp.275-280
• /
• 2003

공간 자료에서 이상점이 존재할 경우 변이도(Variogram)를 추정함에 있어 그 효과를 줄이기 위한 방법으로 로버스트(robust) 변이도를 이용한다. 그러나 이상점이 존재하는 자료분석에서 로버스트 변이도를 사용하기에 앞서 이상점을 수정한 자료를 사용하였을 경우 그 효율성 또한 좋다고 알려져 있다. 본 논문에서는 이상점이 존재하는 자료를 분석함에 있어 기존의 이상점 수정법 및 새로운 이상점 수정법의 효율성을 비교하였다.

• The Korean Journal of Applied Statistics
• /
• v.22 no.6
• /
• pp.1355-1364
• /
• 2009

This work propose a robust test for location parameters in multivariate data based on MVE and MCD with the affine equivariance and the high-breakdown properties. We consider the hypothesis testing satisfying high efficiency and high test power simultaneously to bring in the one-step reweighting procedure upon high-breakdown estimators, which generally suffer from the low efficiency and, as a result, usually used only in the exploratory analysis. Monte Carlo study shows that the suggested method retains nominal significance levels and higher testing power without regard to various population distributions than a Hotelling's $T^2$ test. In an example, a data set containing known outliers does not make an influence toward our proposal, while it renders a Hotelling's $T^2$ useless.

• Journal of the Korean Data and Information Science Society
• /
• v.9 no.2
• /
• pp.345-355
• /
• 1998

The back propagation(BP) algorithm allows multilayer feedforward neural networks to learn input-output mappings from training samples. It iteratively adjusts the network parameters(weights) to minimize the sum of squared approximation errors using a gradient descent technique. However, the mapping acquired through the BP algorithm may be corrupt when errorneous training data are employed. In this paper two types of robust backpropagation algorithms are discussed both from a theoretical point of view and in the case studies of nonlinear regression function estimation and handwritten Korean character recognition. For future research we suggest Bayesian learning approach to neural networks and compare it with two robust backpropagation algorithms.

• 한국지구물리탐사학회:학술대회논문집
• /
• 2005.05a
• /
• pp.273-280
• /
• 2005

Robust magneto-telluric (MT) response function estimators are now in standard use in electromagnetic induction research. Properly devised and applied, these methods can reduce the influence of unusual data (outlier) in the response (electric field) variable, but often not sensitive to exceptional predictor (magnetic field) data, which are termed leverage points. A bounded influence estimator is described which simultaneously limits the influence of both outlier and leverage point, and has proven to consistently yield more reliable MT response function estimates than conventional robust approach. The bounded influence estimator combines a standard robust M-estimator with leverage weighting based on the statistics of the hat matrix diagonal, which is a standard statistical measure of unusual predictors. Further extensions to MT data analysis are proposed, including a establishment of data rejection criterion which minimize the influence of both electric and magnetic outlier in frequency domain based on statistical distribution of electromagnetic field. The rejection scheme made in this study seems to have an effective performance on eliminating extreme data, which is even not removed by BI estimator, in frequency domain. The effectiveness and advantage of these developments are illustrated using real MT data.

• The Korean Journal of Applied Statistics
• /
• v.24 no.5
• /
• pp.931-939
• /
• 2011

In practice, thresholds are determined by the two subjective assessment methods in a generalized pareto distribution of mean extreme function(MEF-graph) or Hill-graph. To remedy the problem of subjectiveness of these methods, we propose an alternative method to determine the threshold based on the robust statistics. We compared the MEF-graph, Hill-graph and our method through VaRs on the Korean stock market data from January 5, 1987 to August 3, 2009. As a result, the VaR based on the proposed method is not much different from the existing methods, and the standard deviation of VaR for our method was the smallest. The results show that our method can be a promising alternative to determine thresholds of the generalized pareto distributions.