Study on Outlier Detection Method in Survival Analysis: Weibull Regression Outlier Model

Journal of Biometrics & Biostatistics

article

Chang Shu¹  Tingting Qin²  Xiaoping Chen³  Ping Yin² 

[1] Hubei Province for the Clinical Medicine Research Center of Hepatic Surgery, Key Laboratory of Organ Transplantation, Ministry of Education and Ministry of Public Health, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology;Department of Epidemiology and Biostatistics and State Key Laboratory of Environment Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology;HUST, Hubei Province for the Clinical Medicine Research Center of Hepatic Surgery;Key Laboratory of Organ Transplantation, Ministry of Education and Ministry of Public Health, Hepatic Surgery Centre at Tongji Hospital, Tongji Medical College

关键词: Survival analysis;    Outlier;    Bayesian method;    Weibulldistribution;    Lasso;   

DOI  :  10.4172/2155-6180.1000410

来源: Hilaris Publisher

【 摘 要 】

Background: This study intends to construct Weibull regression outlier model with outlier and using Bayesianmethod to get parameters estimation and statistical inference. The proposal model may contribute to further thoroughlyand systematically complement and implement of outlier detection methods in survival analysis and fully excavate andutilize the survival data.Method: We construct the Weibull regression outlier model by introducing an n-dimensional shift vector as an outlierindicator to the traditional Weibull regression model. The Bayesian method is used for parameters estimation and MCMCmethod is used for statistical inference. The prior for γ is conditional Laplace distribution and the point estimation ofγ is posterior median. According to confidence interval criterion, the components of γ whose 50% confidence intervalcontained 0 are shrank to 0. Then the nonzero components of γ are supposed to be outliers.Results: The results of simulation study and real example study show that the proposal models are not sensitive tocensor rate of data and the ratio of outlier would slightly influence the accuracy of proposal models. The estimations ofcoefficient of outlier models are robust.Conclusion: The outliers in survival data may contain the new information related to the prognosis of disease whichhas not been discovered yet. By the proposal WROM, we could achieve outlier detection and parameter robust estimationat the same time.

【 授权许可】

Unknown