• Journal of the Korean Data and Information Science Society
• /
• v.26 no.1
• /
• pp.217-227
• /
• 2015

In observational study, handling confounders is a primary issue in measuring treatment effect of interest. Historically, a regression with covariate adjustment (covariate-adjusted regression) has been the typical approach to estimate treatment effect incorporating potential confounders into model. However, ever since the introduction of the propensity score, covariate-adjusted regression has been gradually replaced in medical literatures with various balancing methods based on propensity score. On the other hand, there is only a paucity of researches assessing propensity score methods compared with the covariate-adjusted regression. This paper examined the performance of propensity score methods in estimating risk difference and compare their performance with the covariate-adjusted regression by a Monte Carlo study. The study demonstrated in general the covariate-adjusted regression with variable selection procedure outperformed propensity-score-based methods in terms both of bias and MSE, suggesting that the classical regression method needs to be considered, rather than the propensity score methods, if a performance is a primary concern.

• Communications for Statistical Applications and Methods
• /
• v.19 no.5
• /
• pp.721-729
• /
• 2012

Various methods control the influence of a covariate on a response variable. These methods are analysis of covariance(ANCOVA), RANK ANCOVA, ANOVA of (covariate-adjusted) residuals, and Kruskal-Wallis tests on residuals. Covariate-adjusted residuals are obtained from the overall regression line fit to the entire data set that ignore the treatment levels or factors. It is demonstrated that the methods on covariate-adjusted residuals are only appropriate when the regression lines are parallel and covariate means are equal for all treatments. In this paper, we proposed the new nonparametric method on the ANCOVA model, as applying joint placement in a one-way layout on residuals as described in Chung and Kim (2007). A Monte Carlo simulation study is adapted to compare the power of the proposed procedure with those of the previous procedure.

• The Korean Journal of Applied Statistics
• /
• v.30 no.3
• /
• pp.427-439
• /
• 2017

Quade (1967) proposed RANK ANCOVA, which is a nonparametric method to test differences between treatments when there are covariates. Hwang and Kim (2012) also proposed a joint placement test on covariate-adjusted residuals. In this paper, we proposed a new nonparametric method to control the effect of covariate on a response variable that uses linear statistics on covariate adjusted-residuals. The score function used in the linear statistics was proposed by Jeon and Kim (2016). Monte Carlo simulation is also conducted to compare the empirical powers of the proposed method with previous methods.

• Proceedings of the Korean Statistical Society Conference
• /
• 2002.11a
• /
• pp.197-201
• /
• 2002

The healthy worker effect is an important issue in occupational epidemiology. We proposed a new statistical method to test the relationship between exposure and time to death in the presence of the healthy worker effect. In this study, we considered the healthy worker hire effect to operate as a confounder and the healthy worker survival effect to operate as a confounder and an intermediate variable. The basic idea of the proposed method reflects the length bias-sampling caused by changing one's employment status. Simulation studies were also carried out to compare the proposed method with the Cox proportional hazards models. According to our simulation studies, both the proposed test and the test based on the Cox model having the change of the employment status as a time-dependent covariate seem to be satisfactory at an upper 5% significance level. The Cox models, however, are inadequate with the change, if any, of the employment status as time-independent covariate. The proposed test is superior in power to the test based on the Cox model including the time-dependent employment status.

• The Korean Journal of Applied Statistics
• /
• v.29 no.5
• /
• pp.849-858
• /
• 2016

It is important to control confounding bias when estimating the causal effect of treatment in an observational study. We illustrated that the covariate selection in the causal inference is different from the variable selection in the ANCOVA model. We then investigated the three criteria of covariate selection for controlling confounding bias, which can be used when we have inadequate information to draw a complete causal graph. VanderWeele and Shpitser (2011) proposed one of them and claimed it was better than the other two. We show by example that their criterion also has limitations and some disadvantages. There is no clear winner; however, their criterion is better (if some correction is made on its condition) than the other two because it can remove the confounding bias.

• Survey Research
• /
• v.10 no.3
• /
• pp.21-36
• /
• 2009

This study suggests the applicability of web surveys regarding elections in order to contact a great number of young people. The propensity weighting model was estimated using the demographic variables and the covariate variables collected during the 2007 presidential election surveys. In order to adjust the internet survey to the telephone survey, we used the propensity score method. Propensity score weighting made the internet survey results closer to the telephone survey results. This shows that an internet survey with propensity weighting model is a potential alternative survey method in the prediction of elections.

• Communications for Statistical Applications and Methods
• /
• v.24 no.2
• /
• pp.173-191
• /
• 2017

This study develops a new type of latent class analysis (LCA) in order to explain the associations between one latent variable and several other categorical latent variables. Our model postulates that the prevalence of the latent variable of interest is affected by another latent variable composed of other several latent variables. For the parameter estimation, we propose deterministic annealing EM (DAEM) to deal with local maxima problem in the proposed model. We perform simulation study to demonstrate how DAEM can find the set of parameter estimates at the global maximum of the likelihood over the repeated samples. We apply the proposed LCA model in an investigation of the effect of and joint patterns for drug-using behavior to violent behavior among US high school male students using data from the Youth Risk Behavior Surveillance System 2015. Considering the age of male adolescents as a covariate influencing violent behavior, we identified three classes of violent behavior and three classes of drug-using behavior. We also discovered that the prevalence of violent behavior is affected by the type of drug used for drug-using behavior.

• The Korean Journal of Applied Statistics
• /
• v.35 no.3
• /
• pp.335-346
• /
• 2022

Profile control chart aims to detect a change in the functional relationship of multivariate characteristics in the statistical process control. In monitoring two variables, a linear profile is of interest composed of the intercept and slope of one variable (response variable) against the other (explanatory variable). The previous studies on monitoring of the linear profile mostly assume that the explanatory variables are the same for all profiles. However, there are also cases where they vary depending on profiles. This paper intends to extend the monitoring method to where explanatory variables are different for each profile. We compare the new method's performance through simulation and apply it to monitoring a network intrusion using NSL-KDD data.

• Journal of Korean Society of Transportation
• /
• v.29 no.4
• /
• pp.29-41
• /
• 2011

This study investigates the factor that influences commuting time to work when individuals allocate their time for different types of activities. The commuting time is an important indicator for an individual to determine the residence and choose the means of transportation. The analysis uses the data collected from people who live in Seoul metropolitan area including Seoul, Incheon and Gyeonggi Province, and commute to work and making the simple linkage travel (home-work-home) within the area. For the analysis, the Cox hazard proportional methodology was adopted. The method is known to be well applied without assuming any distribution in case of the dependent variable being continuous. For the covariate, the interaction effect between the space variable of the work place and the variable of transportation has been also included in the model. The commuting time to work has been estimated for both 1) the whole metropolitan area and 2) the separate regions i.e., Seoul, Incheon and Gyeonggi-Do. The result reveals that characteristic variables related to individual, household and travel properties influence the mode of transportation and the time allocated for commuting to work (p<0.01). This study also demonstrates the usefulness of the Cox hazard proportional model. The data used in this study is the actual household travel data surveyed in 2006 in the metropolitan area, and analyzing the survey data in 2010 is currently in progress. Comparison of the two survey data sets seeking any behavioral change is suggested for the future study.

• The Korean Journal of Applied Statistics
• /
• v.27 no.2
• /
• pp.227-237
• /
• 2014

We use the regression model with the indicator variables in the case that we use qualitative variables as some predictor variables in regression analysis. We use the ANCOVA(Analysis of Covariance) model when comparing the response variable among groups while statistically controlling for variation in the response variable caused by a variation in the covariate. DD-plot can be used as a graphical exploratory data analysis tool before the confirmatory data analysis. With the DD-plot, we can discriminate the difference of groups in the regression model with the indicator variables or the ANCOVA model at a glance. Making DD-plot does not demand the statistical model assumption about error terms in regression model. Several examples show the usefulness of DD-plots as a graphical exploratory data analysis tool for the regression analysis.