• Communications for Statistical Applications and Methods
• /
• v.6 no.3
• /
• pp.967-976
• /
• 1999

In this paper we study the Hellinger distance based goodness-of-fit tests that are analogs of likelihood ratio tests. The minimum Hellinger distance estimator (MHDE) in normal models provides an excellent robust alternative to the usual maximum likelihood estimator. Our simulation results show that the Hellinger deviance test (Simpson 1989) based goodness-of-fit test is robust when data contain outliers. The proposed hellinger deviance test(Simpson 1989) is a more direcct method for obtaining robust inferences than an automated outlier screen method used before the likelihood ratio test data analysis.

• 제어로봇시스템학회:학술대회논문집
• /
• 1999.10a
• /
• pp.92-95
• /
• 1999

The design problem of the control system is the ability to synthesize controller that achieve robust stability and robust performance. The paper explains the Finite Inclusions Theorem (FIT) by the procedure namely FIT synthesis. It is developed for synthesizing robustly stabilizing controller for parametrically uncertain system. The fundamental problem in the study of parametrically uncertain system is to determine whether or not all the polynomials in a given family of characteristic polynomials is Hurwitz i.e., all their roots lie in the open left-half plane. By FIT it can prove a polynomial is Hurwitz from only approximate knowledge of the polynomial's phase at finitely many points along the imaginary axis. An example shows the simplicity of using the FIT synthesis to directly search for robust controller of parametrically uncertain system by way of solving a sequence of systems of linear inequalities. The systems of inequalities are solved via the projection method which is an elegantly simple technique fur solving (finite or infinite) systems of convex inequalities in an arbitrary Hilbert space. Results from example show that the controller synthesized by FIT synthesis is better than by H$\sub$$\infty$/ synthesis with parametrically uncertain system as well as satisfied the objectives for a considerably larger range of uncertainty.

• IE interfaces
• /
• v.23 no.2
• /
• pp.118-125
• /
• 2010

The determination of a regression model is important in using statistical designs of experiments. Generally, the exact regression model is not known, and experimenters suppose that a certain model form will be fit. Then an experimental design suitable for that predetermined model form is selected and the experiment is conducted. However, the initially chosen regression model may not be correct, and this can result in undesirable statistical properties. We develop model-robust experimental designs that have stable prediction variance for a family of candidate regression models over a cuboidal region by using genetic algorithms and the desirability function method. We then compare the stability of prediction variance of model-robust experimental designs with those of the 3-level face centered cube. These model-robust experimental designs have moderately high G-efficiencies for all candidate models that the experimenter may potentially wish to fit, and outperform the cuboidal design for the second-order model. The G-efficiencies are provided for the model-robust experimental designs and the face centered cube.

• 한국데이터정보과학회:학술대회논문집
• /
• 2003.10a
• /
• pp.93-100
• /
• 2003

It is necessary to check the curvature of selected covariates in regression diagnostics. There are various graphical methods using residual plots based on least squares fitting. The sensitivity of LS fitting to outliers can distort their residuals, making the identification of the unknown function difficult to impossible. In this paper, we compare combining conditional expectation and residual plots(CERES Plots) between least square fit and robust fits using Huber M-estimator. Robust CERES will be far less distorted than their LS counterparts in the presence of outliers and hence, will be more useful in identifying the unknown function.

• Journal of the Korean Data and Information Science Society
• /
• v.7 no.2
• /
• pp.203-210
• /
• 1996

A linear neural unit with a modified anti-Hebbian learning rule has been shown to be able to optimally fit curves, surfaces, and hypersurfaces by adaptively extracting the minor component of the input data set. In this paper, we study how to use the robust version of this neural fitting method for linear regression analysis. Furthermore, we compare this method with other methods when data set is contaminated by outliers.

• Korean Management Science Review
• /
• v.24 no.2
• /
• pp.73-80
• /
• 2007

Robust Design(RD) is a cost-effective methodology to determine the optimal settings of control factors that make a product performance insensitive to the influence of noise factors. To better facilitate the robust design optimization, a dual response surface approach, which models both the process mean and standard deviation as separate response surfaces, has been successfully accepted by researchers and practitioners. However, the construction of response surface approximations has been limited to problems with only a few variables, mainly due to an excessive number of experimental runs necessary to fit sufficiently accurate models. In this regard, an innovative response surface approach has been proposed to investigate robust design optimization problems with larger number of variables. Response surfaces for process mean and standard deviation are partitioned and estimated based on the multi-level approximation method, which may reduce the number of experimental runs necessary for fitting response surface models to a great extent. The applicability and usefulness of proposed approach have been demonstrated through an illustrative example.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.27 no.9
• /
• pp.1437-1443
• /
• 2003

Most optimization problems do not consider tolerance of design variables and design parameters. Ignorance of these tolerances may not fit for the practical problems and can lead to an unexpected conclusion. That is why we suggest robust optimization considering tolerances in both design variables and problem parameters. Using robust optimization, we designed minimum weight annular finned heat transfer tube subject to constraints on limitation of pressure difference and minimum value of total heat transfer. Consequently, robust optimization satisfies tolerance considered practical problems.

• Communications for Statistical Applications and Methods
• /
• v.14 no.2
• /
• pp.401-411
• /
• 2007

In this paper we propose an influence measure for detecting potentially influential observations using the infinitesimal perturbation and the local influence in the log-linear regression model. Also, we propose a goodness-of-fit measure for variable selection. A real data set are used for illustration.

• Journal of the Korean Data and Information Science Society
• /
• v.15 no.2
• /
• pp.307-316
• /
• 2004

Principal Component Regression(PCR) and Partial Least Squares Regression(PLSR) are the two most popular regression techniques in chemometrics. In the field of chemometrics usually the number of regressor variables greatly exceeds the number of observation. So we have to reduce the number of regressors to avoid the identifiability problem. In this paper we compare PCR and PLSR techniques combined with various robust regression methods including regression depth estimation. We compare the efficiency, goodness-of-fit and robustness of each estimators under several contamination schemes.

• Proceedings of the KSME Conference
• /
• 2003.04a
• /
• pp.170-177
• /
• 2003

This paper proposes a robust method for the Ramberg-Osgood (R-O) fit to accurately estimate elastic-plastic J from engineering fracture mechanics analysis based on deformation plasticity. The proposal is based on engineering stress-strain data to determine the R-O parameters, instead of true stress-strain data. Moreover, for practical applications, the method is given not only for the case when full stress-strain data are available but also for the case when only yield and tensile strengths are available. Reliability of the proposed method for the R-O fit is validated against detailed 3-D Finite Element (FE) analyses for circumferential through-wall cracked pipes under global bending using five different materials, three stainless steels and two ferritic steels. Taking the FE J results based on incremental plasticity using actual stress-strain data as reference, the FE J results based on deformation plasticity using various R-O fits are compared with reference J values. Comparisons show that the proposed R-O fit provides more accurate J values for all cases, compared to existing methods for the R-O fit. Advantages of the proposed R-O fit in practical applications are discussed, together with its accuracy.