• The Korean Journal of Applied Statistics
• /
• v.33 no.5
• /
• pp.569-578
• /
• 2020

By estimating conditional quantile functions of the response, quantile regression (QR) can provide comprehensive information of the relationship between the response and the predictors. In addition, kernel quantile regression (KQR) estimates a nonlinear conditional quantile function in reproducing kernel Hilbert spaces generated by a positive definite kernel function. However, it is infeasible to use the KQR in analysing a massive data due to the limitations of computer primary memory. We propose a divide and conquer based KQR (DC-KQR) method to overcome such a limitation. The proposed DC-KQR divides the entire data into a few subsets, then applies the KQR onto each subsets and derives a final estimator by aggregating all results from subsets. Simulation studies are presented to demonstrate the satisfactory performance of the proposed method.

• Journal of Korea Water Resources Association
• /
• v.45 no.8
• /
• pp.815-826
• /
• 2012

The underestimating trend using existing ordinary regression (OR) based trend analysis has been a well-known problem. The existing OR method based on least squares approximate the conditional mean of the response variable given certain values of the time t, and the usual assumption of the OR method is normality, that is the distribution of data are not dissimilar form a normal distribution. In this regard, this study proposed a quantile regression that aims at estimating either the conditional median or other quantiles of the response variable. This study assess trend in annual daily maximum rainfall series over 64 weather stations through both in OR and QR approach. The QR method indicates that 47 stations out of 67 weather stations are a strong upward trend at 5% significance level while OR method identifies a significant trend only at 13 stations. This is mainly because the OR method is estimating the condition mean of the response variable. Unlike the OR method, the QR method allows us flexibly to detect the trends since the OR is designed to estimate conditional quantiles of the response variable. The proposed QR method can be effectively applied to estimate hydrologic trend for either non-normal data or skewed data.

• The Korean Journal of Applied Statistics
• /
• v.25 no.1
• /
• pp.197-203
• /
• 2012

In fitting a regression model, we often encounter data sets which do not follow Gaussian distribution and/or do not have equal variance. In this case estimation of the conditional density of a response variable at a given design point is hardly solved by a standard least squares method. To solve this problem, we propose a simple method to estimate the distribution of the fitted vales under heteroscedasticity using the idea of quantile regression and the histogram techniques. Application of this method to a real data sets is given.

• The Korean Journal of Applied Statistics
• /
• v.29 no.7
• /
• pp.1361-1371
• /
• 2016

Quantile regression provides a variety of useful statistical information to examine how covariates influence the conditional quantile functions of a response variable. However, traditional quantile regression (which assume a linear model) is not appropriate when the relationship between the response and the covariates is a nonlinear. It is also necessary to conduct variable selection for high dimensional data or strongly correlated covariates. In this paper, we propose a penalized quantile regression tree model. The split rule of the proposed method is based on residual analysis, which has a negligible bias to select a split variable and reasonable computational cost. A simulation study and real data analysis are presented to demonstrate the satisfactory performance and usefulness of the proposed method.

• Communications for Statistical Applications and Methods
• /
• v.6 no.1
• /
• pp.25-32
• /
• 1999

A single outlier can influence on the least squares estimators and can invalidate analysis based on these estimators. The Cook's statistic has been introduced to measure influence of individual data point on parameter estimation and the quantile of the F distribution is recommended as a reference value. but in practice subjective judgement is applied in the choice of appropriate quantile. A simple reference value is introduced in this paper which is developed by approximating conditional quantities of Cook's measure. The performance of the proposed criterion is evaluated through analysis of real data set.

• The Korean Journal of Applied Statistics
• /
• v.29 no.6
• /
• pp.1095-1106
• /
• 2016

The quantile regression method proposed by Koenker et al. (1978) focuses on conditional quantiles given by independent variables, and analyzes the relationship between response variable and independent variables at the given quantile. Considering the linear programming used for the estimation of quantile regression coefficients, the model fitting job might be difficult when large data are introduced for analysis. Therefore, dimension reduction (or variable selection) could be a good solution for the quantile regression of large data sets. Regression tree methods are applied to a variable selection for quantile regression in this paper. Real data of Korea Baseball Organization (KBO) players are analyzed following the variable selection approach based on the regression tree. Analysis result shows that a few important variables are selected, which are also meaningful for the given quantiles of salary data of the baseball players.

• The Journal of Asian Finance, Economics and Business
• /
• v.7 no.8
• /
• pp.161-173
• /
• 2020

This study examines the nonlinear relationship between financial development and economic growth in Pakistan using the threshold regression model for the period 1980-2017. We also employed quantile regression with 0.25, 0.50, and 0.75 quantiles of conditional distribution. The quantile regression is based on minimizing of sum of squared residuals. The result indicates that economic growth responds positively to financial development when the level of financial development surpasses the threshold value of 0.151. However, when financial development lies below the threshold value (that is, 0.151), its impact on economic growth is negative. Thus, when financial development of Pakistan surpasses the threshold level, it contributes more towards economic growth since greater level of financial development contributes more to boosts economic growth. This finding reveals that economic growth reacts differently to financial development, and the relationship between financial development and economic growth is U-shaped in Pakistan. Among the other variables, physical capital, labor force, and government expenditure exert a positive effect on economic growth. Furthermore, inflation rate and trade openness have an insignificant impact on economic growth. The results of quantile regression also confirm the non-linear relationship between financial development and economic growth in Pakistan. The finding of this study suggests revamping of financial sector policies in Pakistan.

• The Korean Journal of Applied Statistics
• /
• v.29 no.7
• /
• pp.1201-1212
• /
• 2016

Value at Risk (VaR) for market risk management is a favorite method used by financial companies; however, there are some problems that cannot be explained for the amount of loss when a specific investment fails. Conditional Tail Expectation (CTE) is an alternative risk measure defined as the conditional expectation exceeded VaR. Multivariate loss rates are transformed into a univariate distribution in real financial markets in order to obtain CTE for some portfolio as well as to estimate CTE. We propose multivariate CTEs using multivariate quantile vectors. A relationship among multivariate CTEs is also derived by extending univariate CTEs. Multivariate CTEs are obtained from bivariate and trivariate normal distributions; in addition, relationships among multivariate CTEs are also explored. We then discuss the extensibility to high dimension as well as illustrate some examples. Multivariate CTEs (using variance-covariance matrix and multivariate quantile vector) are found to have smaller values than CTEs transformed to univariate. Therefore, it can be concluded that the proposed multivariate CTEs provides smaller estimates that represent less risk than others and that a drastic investment using this CTE is also possible when a diversified investment strategy includes many companies in a portfolio.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.15 no.3
• /
• pp.118-128
• /
• 2016

As the rural areas have experienced the population reduction and the aging, the service level of public transit decreases. This study analyzes the effecting factor to user satisfaction of demand response transport(DRT) as alternative to rural public transit by the quantile regression that aims at estimating either the conditional median or other quantiles of the response variable. Jeonbuk Province tested DRT operations in Dongsang of Wanju County and Sannae of Jeongup City each in 2015. The user DRT satisfaction of Wanju was higher than one of Jeongup in basic statistics analysis. The difference in satisfaction between higher quantile and lower quntile of Wanju is smaller than one of Jeongupy as a result of quantile regression analysis. Also, Wanju DRT continues the second test operation of DRT as satisfaction from Ordinary Least Squares(OLS) close to higher satisfaction quantile.

• Journal of the Korean Data and Information Science Society
• /
• v.28 no.3
• /
• pp.533-545
• /
• 2017

Quantile regression models provide a variety of useful statistical information by estimating the conditional quantile function of the response variable. However, the traditional linear quantile regression model can lead to the distorted and incorrect results when analysing real data having a nonlinear relationship between the explanatory variables and the response variables. Furthermore, as the complexity of the data increases, it is required to analyse multiple response variables simultaneously with more sophisticated interpretations. For such reasons, we propose a multivariate quantile regression tree model. In this paper, a new split variable selection algorithm is suggested for a multivariate regression tree model. This algorithm can select the split variable more accurately than the previous method without significant selection bias. We investigate the performance of our proposed method with both simulation and real data studies.