• Communications for Statistical Applications and Methods
• /
• v.14 no.1
• /
• pp.155-167
• /
• 2007

Knowing the time of the process change could lead to quicker identification of the responsible special cause and less process down time, and it could help to reduce the probability of incorrectly identifying the special cause. In this paper, we propose the maximum likelihood estimator (MLE) for the process change point when a control chart is used in monitoring the mean of a process in which the observations can be modeled as an AR(1) process plus an additional random error. The performance of the proposed MLE is compared to the performance of the built-in estimator when they are used in EWMA charts based on the residuals. The results show that the proposed MLE provides good performance in terms of both accuracy and precision of the estimator.

• Journal of the Korean Data and Information Science Society
• /
• v.16 no.2
• /
• pp.173-183
• /
• 2005

This paper studies quick detection of variance change point for iid data. For development of sensitive and adaptive variance change point detector, moving variance ratio is employed as a variance ratio estimator. It is shown that selection of tuning parameters of detector, (i.e., information and lag tuning parameters) is critical for detector to achieve desirable sensitivity and adaptiveness. Interestingly our simulation result reveals limitations of the commonly used change ratio against the previous day. Our results will provide useful insight when the detector is applied to time series data.

• Journal of the Korean Data and Information Science Society
• /
• v.16 no.2
• /
• pp.289-300
• /
• 2005

In this article quick variance change point (VCP) detection problem for time series is considered. For this variance VCP detector equipped with tuning parameters is proposed. A major tool for the detector is moving variance ratio (MVR) which monitors variance change of a given time series. Tuning process of detector is investigated via simulation, which shows that tuning parameters are critical in achieving sensitivity and adaptiveness of detector.

• The Korean Journal of Applied Statistics
• /
• v.25 no.6
• /
• pp.999-1008
• /
• 2012

A Bayesian multiple change-point model for small data is proposed for multivariate means and is an extension of the univariate case of Cheon and Yu (2012). The proposed model requires data from a multivariate noncentral $t$-distribution and conjugate priors for the distributional parameters. We apply the Metropolis-Hastings-within-Gibbs Sampling algorithm to the proposed model to detecte multiple change-points. The performance of our proposed algorithm has been investigated on simulated and real dataset, Hanwoo fat content bivariate data.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2013.05a
• /
• pp.659-660
• /
• 2013

• Proceedings of the Korean Statistical Society Conference
• /
• 2002.11a
• /
• pp.103-108
• /
• 2002

We consider an estimation of discontinuous variance function in nonparametric heteroscedastic random design regression model. We first propose estimators of a change point and jump size in variance function and then construct an estimator of entire variance function. We examine the rates of convergence of these estimators and give results on their asymptotics. Numerical work reveals that the effectiveness of change point analysis in variance function estimation is quite significant.

• Journal of the Korean Data and Information Science Society
• /
• v.14 no.3
• /
• pp.641-651
• /
• 2003

This study suggests the hybrid models for interest rate forecasting using structural changes (or change points). The basic concept of this proposed model is to obtain significant intervals caused by change points, to identify them as the change-point groups, and to reflect them in interest rate forecasting. The model is composed of three phases. The first phase is to detect successive structural changes in the U. S. Treasury bill rate dataset. The second phase is to forecast the change-point groups with data mining classifiers. The final phase is to forecast interest rates with backpropagation neural networks (BPN). Based on this structure, we propose three hybrid models in terms of data mining classifier: (1) multivariate discriminant analysis (MDA)-supported model, (2) case-based reasoning (CBR)-supported model, and (3) BPN-supported model. Subsequently, we compare these models with a neural network model alone and, in addition, determine which of three classifiers (MDA, CBR and BPN) can perform better. For interest rate forecasting, this study then examines the prediction ability of hybrid models to reflect the structural change.

• Communications for Statistical Applications and Methods
• /
• v.24 no.3
• /
• pp.211-226
• /
• 2017

The chain-ladder method, for which run-off data is employed is popularly used in the rate-adjustment and loss-reserving practices of non-life-insurance and health-insurance companies. The method is applicable when the underlying assumption of a consistent development pattern is in regards to a cumulative loss payment after the occurrence of an insurance event. In this study, a modified chain-ladder algorithm is proposed for when the assumption is considered to be only partially appropriate for the given run-off data. The concept of a structural-change point in the run-off data and its reflection in the estimation of unpaid loss amounts are discussed with numerical illustrations. Experience data from private health insurance coverage in Korea were analyzed based on the suggested method. The performance in estimation of loss reserve was also compared with traditional approaches. We present evidence in this paper that shows that a reflection of a structural-change point in the chain-ladder method can improve the risk management of the relevant insurance products. The suggested method is expected to be utilized easily in actuarial practice as the algorithm is straightforward.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.6
• /
• pp.1225-1238
• /
• 2015

In this research, we fit various survival models and conduct tests and estimation for the hazard change-point with the rectal cancer data. By the log-rank tests, at significance level ${\alpha}=0.10$, survival functions are significantly different according to the uniporter of glucose (GLUT1), clinical stage (cstage) and pathologic stage (ypstage). From the Cox proportional hazard model, the most significant covariates are GLUT1 and ypstage. Assuming that the rectal cancer data follows the exponential distribution, we estimate one hazard change-point using Matthews and Farewell (1982), Henderson (1990) and Loader (1991) methods.

• Communications for Statistical Applications and Methods
• /
• v.3 no.2
• /
• pp.275-282
• /
• 1996

A life distribution F with survival function $\overline{F}$=1-F, finite mean $\mu$ and mean residual life m(t) is said to be NBUE(NWUE) if m(t)$\leq$($\geq$) .$\mu$ for t$\geq$0. This NBUE property can equivalently be characterized by the fact that $\varphi$(u)$\geq$($\leq$)u for 0$\leq$u$\leq$1, where $\varphi$(u) is the scaled total-time-on test transform of F. A generalization of the NBUE properties is that there is a value of p such that $\varphi$(u)\geq.u$ for 0$\leq$u$\leq$p and $\varphi$(u)\leq$$\leq$u$\leq$1, or vice versa. This means that we have a trend change in the NBUE property. In this paper we point out an error of Klefsjo's paper (1988). He erroneously takes advantage of trend change point of failure rate to calculate the empirical test size and power in lognormal distribution. We solves the trend change point of mean residual lifetime and recalculate the empirical test size and power of Klefsjo (1988) in mocensoring case.