• Communications for Statistical Applications and Methods
• /
• v.10 no.1
• /
• pp.167-175
• /
• 2003

In this paper we develop a robust procedure to estimate regression coefficients for a linear model with censored and truncated data based on simplicial regression depth. Simplicial depth of a point is defined as the proportion of data simplices containing it. This simplicial depth can be extended to regression problem with censored and truncated data. Any line can be given a depth and the deepest regression line is the line with the maximum simplicial regression depth. We show how the proposed regression performs through analyzing AIDS incubation data.

• 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.

• The Journal of the Korea Contents Association
• /
• v.15 no.1
• /
• pp.475-481
• /
• 2015

We use polynomial regression instead of linear regression if there is a nonlinear relation between a dependent variable and independent variables in a regression analysis. The performance of polynomial regression, however, may deteriorate because of the correlation caused by the power terms of independent variables. We present a polynomial regression model for the numerical inversion of PGF and show that polynomial regression results in the deterioration of the estimation of the coefficients. We apply principal components regression to the polynomial regression model and show that principal components regression dramatically improves the performance of the parameter estimation.

• Korean Journal of Remote Sensing
• /
• v.38 no.4
• /
• pp.327-341
• /
• 2022

Cloud removal is often required to construct time-series sets of optical images for environmental monitoring. In regression-based cloud removal, the selection of an appropriate regression model and the impact analysis of the input images significantly affect the prediction performance. This study evaluates the potential of Gaussian process (GP) regression for cloud removal and also analyzes the effects of cloud-free optical images and spectral bands on prediction performance. Unlike other machine learning-based regression models, GP regression provides uncertainty information and automatically optimizes hyperparameters. An experiment using Sentinel-2 multi-spectral images was conducted for cloud removal in the two agricultural regions. The prediction performance of GP regression was compared with that of random forest (RF) regression. Various combinations of input images and multi-spectral bands were considered for quantitative evaluations. The experimental results showed that using multi-temporal images with multi-spectral bands as inputs achieved the best prediction accuracy. Highly correlated adjacent multi-spectral bands and temporally correlated multi-temporal images resulted in an improved prediction accuracy. The prediction performance of GP regression was significantly improved in predicting the near-infrared band compared to that of RF regression. Estimating the distribution function of input data in GP regression could reflect the variations in the considered spectral band with a broader range. In particular, GP regression was superior to RF regression for reproducing structural patterns at both sites in terms of structural similarity. In addition, uncertainty information provided by GP regression showed a reasonable similarity to prediction errors for some sub-areas, indicating that uncertainty estimates may be used to measure the prediction result quality. These findings suggest that GP regression could be beneficial for cloud removal and optical image reconstruction. In addition, the impact analysis results of the input images provide guidelines for selecting optimal images for regression-based cloud removal.

• Communications for Statistical Applications and Methods
• /
• v.8 no.1
• /
• pp.15-27
• /
• 2001

One proposal is made for constructing nonparametric estimator of slope parameters in a regression model under symmetric error distributions. This estimator is based on the use of idea of Johns for estimating the center of the symmetric distribution together with the idea of regression quantiles and regression trimmed mean. This nonparametric estimator and some other L-estimators are studied by Monte Carlo.

• The Korean Journal of Applied Statistics
• /
• v.20 no.2
• /
• pp.313-322
• /
• 2007

Several robust censored depth regression methods are compared under contamination. Park and Hwang(2003) suggested a way to circumvent the censoring issue by incorporating Kaplan-Meier type weight in halfspace regression depth and Park(2003) used a similar technique to simplicial regression depth. Hubert et al. (2001) suggested a high breakdown point regression depth based on projection called rcent. A new method to implement censoring in rcent is suggested and compared with two precedents under various contamination and censoring schemes.

• Communications for Statistical Applications and Methods
• /
• v.9 no.2
• /
• pp.521-531
• /
• 2002

Various diagnostic techniques for identifying influential observations are mostly based on the deletion of a single observation. While such techniques can satisfactorily identify influential observations in many cases, they will not always be successful because of some mask effect. It is necessary, therefore, to develop techniques that examine the potentially influential effects of a subset of observations. The partial regression plots can be used to examine an influential observation for a single parameter in multiple linear regression. However, it is often desirable to detect influential observations for a subset of regression parameters when interest centers on a selected subset of independent variables. Thus, we propose a diagnostic measure which deals with detecting influential observations on a subset of regression parameters. In this paper, we propose a measure M, which can be effectively used for the detection of influential observations on a subset of regression parameters in multiple linear regression. An illustrated example is given to show how we can use the new measure M to identify influential observations on a subset of regression parameters.

• Journal of the Korean Data and Information Science Society
• /
• v.13 no.1
• /
• pp.157-164
• /
• 2002

In this paper, we consider the strong consistency of the regression quantiles estimators for the nonlinear regression models when dependent variables are subject to censoring, and provide the sufficient conditions which ensure the strong consistency of proposed estimators of the censored regression models. one example is given to illustrate the application of the main result.

• Communications for Statistical Applications and Methods
• /
• v.4 no.1
• /
• pp.21-32
• /
• 1997

Robust regression based on M-estimator reduces and/or bounds the influence of outliers in the y-direction only. Therefore, when several influential observations exist, diagnostics in the robust regression is required in order to detect them. In this paper, we propose influence diagnostics in the robust regression based on M-estimator and its one-step version. Noting that M-estimator can be obtained through iterative weighted least squares regression by using internal weights, we apply the weighted least squares (WLS) regression diagnostics to robust regression.

• Communications for Statistical Applications and Methods
• /
• v.7 no.3
• /
• pp.633-641
• /
• 2000

In this paper, we propose a method to identify multiple outliers in regression analysis with only assumption of smoothness on the regression function. Our method uses single-linkage clustering algorithm and Projection Pursuit Regression (PPR). It was compared with existing methods using several simulated and real examples and turned out to be very useful in regression problem with the regression function which is far from linear.