• 한국통계학회:학술대회논문집
• /
• 한국통계학회 2002년도 춘계 학술발표회 논문집
• /
• pp.183-188
• /
• 2002

Multiple imputation, proposed by Rubin, is a procedure for handling missing data. One of the attractive parts of multiple imputation is the simplicity of the variance estimation formula. Because of the simplicity, it has been often abused and misused beyond its original prescription. This paper provides the bias of the multiple imputation variance estimator for a linear point estimator and discusses when the bias can be safely neglected.

• 응용통계연구
• /
• 제32권1호
• /
• pp.83-97
• /
• 2019

사회가 발전함에 따라 이용자의 다양한 분석 요구에 대응하기 위해 개인 단위로 구성된 마이크로데이터 제공이 증가했다. 나아가 센서스, 행정자료와 같은 전수자료를 마이크로데이터 형태로 제공받아 연구하고자 하는 요구 역시 커지고 있다. 정책결정, 학술목적 등을 위한 마이크로데이터 분석은 가치 창출 측면에서 대단히 바람직하다. 하지만 자료 유용성이 확보된 마이크로데이터 제공은 개인정보가 노출될 가능성이라는 위험을 가질 수 밖에 없다. 이에, 자료의 유용성을 확보하면서 개인정보보호를 보장할 수 있는 여러 방법들이 고려되어 왔다. 이러한 방법 중 하나로 재현자료(synthetic data)를 생성해서 활용하는 방법이 연구되어 왔다. 본 논문은 재현자료 생성과 관련된 방법론 및 주의사항을 소개하여, 재현자료의 이해를 도모하고자 한다. 이를 위해 재현자료 작성에 필수적인 다중대체, 베이지안 예측 모형 및 베이지안 붓스트랩 등의 개념들을 먼저 설명하고, 완전 재현자료 및 부분 재현자료에 대해 살펴본다. 특히, 재현자료 작성을 심도 깊이 이해하기 위해 순차회귀 다중대체(sequential regression multivariate imputation)를 이용해 경시적(longitudinal) 자료를 재현자료로 작성하는 구체적 사례를 살펴본다.

• 응용통계연구
• /
• 제32권6호
• /
• pp.889-898
• /
• 2019

항목 무응답(item missing)이 발생한 설문조사에서 결측이 포함된 변수에 이상치(outlier)의 존재와 다른 설문문항 항목과의 논리적 한계(boundary) 조건들이 유의미하다면 결측치 대체문제는 매우 복잡해진다. 한계가 있는 결측값들을 포함한 변수에 이상치가 존재하는 경우, 기존의 회귀분석에 근거한 결측치 대체방법은 편향된 대체값 그리고 한계를 만족하지 않은 대체값을 제시할 가능성이 있다. 이에 본 논문은 회귀모형에 기반을 두고 결측치들을 대체를 함에 있어 이상치와 논리적 한계조건이 자료에 존재하는 경우, 다양한 로버스트 회귀모형과 다중대체 방법의 조합을 통해 해결점을 모색하고자 한다. 이를 위해 이들 방법들의 최적의 조합을 다양한 시나리오별로 모의실험을 통하여 찾아보고 이에 대하여 논의하였다.

• Communications for Statistical Applications and Methods
• /
• 제25권2호
• /
• pp.159-172
• /
• 2018

In many epidemiological studies, missing values in the outcome arise due to censoring. Such censoring is what makes survival analysis special and differentiated from other analytical methods. There are many methods that deal with censored data in survival analysis. However, few studies have dealt with missing covariates in survival data. Furthermore, studies dealing with missing covariates are rare when data are clustered. In this paper, we conducted a simulation study to compare results of several missing data methods when data had clustered multi-structured type with missing covariates. In this study, we modeled unknown baseline hazard and frailty with Bayesian B-Spline to obtain more smooth and accurate estimates. We also used prior information to achieve more accurate results. We assumed the missing mechanism as MAR. We compared the performance of five different missing data techniques and compared these results through simulation studies. We also presented results from a Multi-Center study of Korean IBD patients with Crohn's disease(Lee et al., Journal of the Korean Society of Coloproctology, 28, 188-194, 2012).

• 한국통계학회:학술대회논문집
• /
• 한국통계학회 2000년도 추계학술발표회 논문집
• /
• pp.51-56
• /
• 2000

When the standard inference is to be used with complete data and nonresponse is ignorable, then multiple imputations should be created as repetitions under a Bayesian normal model. Many Bayesian models besides the normal, however, approximately yield the standard inference with complete data and thus many such models can be used to create proper imputations. We consider the Bayesian bootstrap (BB) application.

• Journal of Gastric Cancer
• /
• 제14권4호
• /
• pp.259-265
• /
• 2014

Purpose: Survival analysis of gastric cancer patients requires knowledge about factors that affect survival time. This paper attempted to analyze the survival of patients with incomplete registered data by using imputation methods. Materials and Methods: Three missing data imputation methods, including regression, expectation maximization algorithm, and multiple imputation (MI) using Monte Carlo Markov Chain methods, were applied to the data of cancer patients referred to the cancer institute at Imam Khomeini Hospital in Tehran in 2003 to 2008. The data included demographic variables, survival times, and censored variable of 471 patients with gastric cancer. After using imputation methods to account for missing covariate data, the data were analyzed using a Cox regression model and the results were compared. Results: The mean patient survival time after diagnosis was $49.1{\pm}4.4$ months. In the complete case analysis, which used information from 100 of the 471 patients, very wide and uninformative confidence intervals were obtained for the chemotherapy and surgery hazard ratios (HRs). However, after imputation, the maximum confidence interval widths for the chemotherapy and surgery HRs were 8.470 and 0.806, respectively. The minimum width corresponded with MI. Furthermore, the minimum Bayesian and Akaike information criteria values correlated with MI (-821.236 and -827.866, respectively). Conclusions: Missing value imputation increased the estimate precision and accuracy. In addition, MI yielded better results when compared with the expectation maximization algorithm and regression simple imputation methods.

• Journal of the Korean Data and Information Science Society
• /
• 제27권1호
• /
• pp.217-223
• /
• 2016

Outcome misclassification is widespread in classification problems, but methods to account for it are rarely used. In this paper, the problem of inference with misclassified multinomial logit data with a large number of multinomial parameters is addressed. We have had a significant swell of interest in the development of novel methods to infer misclassified data. One simulation study is shown regarding how seriously misclassification issue occurs if the number of categories increase. Then, using the group lasso regression, we will show how the best model should be fitted for that kind of multinomial regression problems comprehensively.