竞争风险数据中累积发生率置信区间的估计研究
|
摘要
目的累积发生率(CIF)是医学随访中竞争风险型数据重要的描述性指标,CIF置信区间(CI)可以描述其可信程度,然而经典CIF的CI上下限可能出现越界情形,本文研究了基于5种不同转换的CI估计法及其性能。方法将CIF分别基于线性(经典)、对数、双对数、反正弦平方根以及逻辑转换构造CI形式。通过模拟研究,结合方差分析技术综合评价5种CI各自错误覆盖率的平均偏差。结果模拟结果显示线性和反正弦平方根转换均有较大的正数偏差,对数转换则易出现波动,逻辑转换有最小负数偏差,只有双对数转换偏差最接近于期望常数0。结论结合模拟结果和实际例子,线性和对数转换CI容易过宽且表现不稳定,无法克服出现越界异常,反正弦平方根和逻辑转换则轻微波动,只有双对数转换表现最为稳健可靠。
Objective The cumulative incidence function( CIF) is an important descriptive indicator for competing risk data in medical follow-up study. However,the upper and lower limits of the classic confidence interval( CI) of CIF may be exclusive the boundaries. In this paper,the CI estimators based on five different transformations and their performances are studied.Methods The CIs of CIF are constructed based on the linear( classical),log,log(-log),arcsine and logit transformation,respectively. Through the simulation study,the average deviations of the false coverage probabilities for all CIs are comprehensively investigated by the ANOVA technology. Results The simulation results showthat the CIs based on linear and arcsine transformation have a large positive deviation. Log transformation is prone to fluctuations and has a minimum negative deviation,only log(-log) transformation is closest to the expected constant 0,and most robust and reliable. Conclusion Combined with the simulation results and example,CIs base on linear and log transformation are easy to have wide range and unstable performance,and can not overcome the bounds being negative or above 1; the arcsine and logit is slightly fluctuated,but their performances are relatively balanced; only performance of log(-log) is the most robust and reliable.
Objective The cumulative incidence function( CIF) is an important descriptive indicator for competing risk data in medical follow-up study. However,the upper and lower limits of the classic confidence interval( CI) of CIF may be exclusive the boundaries. In this paper,the CI estimators based on five different transformations and their performances are studied.Methods The CIs of CIF are constructed based on the linear( classical),log,log(-log),arcsine and logit transformation,respectively. Through the simulation study,the average deviations of the false coverage probabilities for all CIs are comprehensively investigated by the ANOVA technology. Results The simulation results showthat the CIs based on linear and arcsine transformation have a large positive deviation. Log transformation is prone to fluctuations and has a minimum negative deviation,only log(-log) transformation is closest to the expected constant 0,and most robust and reliable. Conclusion Combined with the simulation results and example,CIs base on linear and log transformation are easy to have wide range and unstable performance,and can not overcome the bounds being negative or above 1; the arcsine and logit is slightly fluctuated,but their performances are relatively balanced; only performance of log(-log) is the most robust and reliable.
引文
[1]Lau B,Cole SR,Gange SJ.Competing risk regression models for epidemiologic data.Am J Epidemiol,2009,170(2):244-256.
[2]Austin PC,Lee DS,Fine JP.Introduction to the Analysis of Survival Data in the Presence of Competing Risks.Circulation,2016,133(6):601-609.
[3]杨召,王少明,粱赫,等.竞争风险型数据统计分析理论研究进展.中国卫生统计,2016,33(6):1088-1091.
[4]卢梓航,周立志,韩栋,等.竞争风险型数据的统计推断处理及应用.现代预防医学,2013,40(5):804-807.
[5]陈征,Nakamura T.基于竞争风险理论和概要型数据的病死率估计模型.中国卫生统计,2010,27(3):249-252.
[6]Kalbfleisch JD,Prentice RL.The Statistical Analysis of Failure Time Data.NewYork:Wiley,2002.
[7]Hong Y,Meeher WQ.Confidence interval procedures for system reliabilityand applications to competing risks models.Lifetime Data A-nal,2014,20(2):161-184.
[8]陈金宝,邱李斌,王北琪,等.固定点处组间生存率比较的统计检验法.中华流行病学杂志,2015,36(2):186-188.
[9]项永兵,高玉堂,金凡,等.生存率置信区间的五种估计方法.中华流行病学杂,1995,16(5):306-309.
[10]Choudhury JB.Non-parametric confidence interval estimation for competing risks analysis:application to contraceptive data.Stat M ed,2002,21(8):1129-1144.
[11]Aalen O.Nonparametric estimation of partial transition probabilities in multiple decrement models.Ann Stat,1978,6(3):534-545.
[12]Beyersmann J,Latouche A,Buchholz A,et al.Simulating competing risks data in survival analysis.Stat M ed,2009,28(6):956-971.
[13]Klein JP,Logan B,Harhoff M,et a1.Analyzing survival curves at a fixed point in time.Stat M ed,2007,26(24):4505-4519.
[14]Pintilie M.Competing Risks:A Practical Perspective.England:John Wiley&Sons,2006.
[2]Austin PC,Lee DS,Fine JP.Introduction to the Analysis of Survival Data in the Presence of Competing Risks.Circulation,2016,133(6):601-609.
[3]杨召,王少明,粱赫,等.竞争风险型数据统计分析理论研究进展.中国卫生统计,2016,33(6):1088-1091.
[4]卢梓航,周立志,韩栋,等.竞争风险型数据的统计推断处理及应用.现代预防医学,2013,40(5):804-807.
[5]陈征,Nakamura T.基于竞争风险理论和概要型数据的病死率估计模型.中国卫生统计,2010,27(3):249-252.
[6]Kalbfleisch JD,Prentice RL.The Statistical Analysis of Failure Time Data.NewYork:Wiley,2002.
[7]Hong Y,Meeher WQ.Confidence interval procedures for system reliabilityand applications to competing risks models.Lifetime Data A-nal,2014,20(2):161-184.
[8]陈金宝,邱李斌,王北琪,等.固定点处组间生存率比较的统计检验法.中华流行病学杂志,2015,36(2):186-188.
[9]项永兵,高玉堂,金凡,等.生存率置信区间的五种估计方法.中华流行病学杂,1995,16(5):306-309.
[10]Choudhury JB.Non-parametric confidence interval estimation for competing risks analysis:application to contraceptive data.Stat M ed,2002,21(8):1129-1144.
[11]Aalen O.Nonparametric estimation of partial transition probabilities in multiple decrement models.Ann Stat,1978,6(3):534-545.
[12]Beyersmann J,Latouche A,Buchholz A,et al.Simulating competing risks data in survival analysis.Stat M ed,2009,28(6):956-971.
[13]Klein JP,Logan B,Harhoff M,et a1.Analyzing survival curves at a fixed point in time.Stat M ed,2007,26(24):4505-4519.
[14]Pintilie M.Competing Risks:A Practical Perspective.England:John Wiley&Sons,2006.