多重填补法和多水平模型在纵向随访数据中的应用

吴秋红; 张丕德; 周国茂; 罗珍胄

doi:10.16462/j.cnki.zhjbkz.2016.07.021

多重填补法和多水平模型在纵向随访数据中的应用

doi: 10.16462/j.cnki.zhjbkz.2016.07.021

1. 深圳市南山区慢性病防治院皮肤性病防治科, 广东深圳 518054;
2. 广东药学院公共卫生学院卫生统计学系, 广东广州 510310

详细信息

作者简介:
吴秋红(1984-),女,广东梅州人,主管医师,硕士。主要研究方向:性病流行病学研究与数据统计分析工作。

中图分类号: R32
计量
- 文章访问数: 523
- HTML全文浏览量: 80
- PDF下载量: 55
- 被引次数: 0
出版历程
- 收稿日期: 2015-07-14
- 修回日期: 2016-01-15

The application of multiple imputation and multilevel model in longitudinal follow-up data

1. Department of Skin Venereal Disease Prevention and Control, Nanshan District Hospital for Chronic Disease Prevention and Treatment, Shenzhen 518054, China;
2. Department of Health Statistics, School of Public Health, Guangdong Pharmaceutical University, Guangzhou 510310, China

摘要

摘要: 目的探讨在纵向随访数据中如何处理缺失值和相关性，充分利用所收集到的数据来反映研究总体。方法先模拟产生纵向完整数据集和缺失数据集，然后用多重填补法（multiple imputation methods，MI）和多水平模型（multilevel model，MLM）来处理，再用随机区组方差分析比较各组的差异，最后用实例验证。结果不同缺失类型和不同缺失比例的数据集所得结果一致：基于MI的MLM所得的偏差比MLM小，且随着填补次数的增多而有所减小；偏差随着缺失率的增大而增加，样本量大的结果更稳定。实例分析也验证了模拟的结果。结论用多重填补法和多水平模型共同处理纵向随访数据可以提高结果的准确性和精确性。
- 纵向研究 /
- 模型,统计学 /
- 随访研究
Abstract: Objective To discuss how to deal with missing data and intra-class correlation in longitudinal follow-up data, and fully use the collected data to reflect the overall information.Methods Firstly, longitudinal complete data and missing data were simulated; Secondly, apply multiple imputation methods(MI) and multilevel model(MLM) were applied to deal with these data sets; Moreover, randomized block analyze of variance was used to analysis each data sets; Finally, an example was used to validate the simulation results.Results The results of different missing type coincided with those of different missing proportion: the deviation of MI with MLM was less than simple MLM. The more filling times, the smaller the deviation. As the missing rate increased, deviation of all methods became larger. The results got more stable in the present of large sample size. The simulation results were also verified by example analysis.Conclusions MI with MLM can provide effective and reasonable results in the longitudinal follow-up data.
- Longitudinal study /
- Model,Statistical /
- Follow-up studies

HTML全文

参考文献(18)

Fitzmaurice G,Laird N,Ware J. Applied longitudinal analysis [M]. New York: John Wiley & Sons, 2004.

王济川,谢海义,姜宝法. 多层统计分析模型:方法与应用 [M]. 北京:高等教育出版社, 2008.

吴秋红,张裕青,李国平,等. 不同模型处理纵向缺失数据的模拟研究及应用 [J]. 中国卫生统计, 2013,30(6):855-858

,861.

李丽霞,周舒冬,张敏,等. 多水平模型和潜变量增长曲线模型在纵向数据分析中的应用及比较 [J]. 中华流行病学杂志, 2014,35(6):741-744.

武海滨,胡如英,方乐,等. 浙江省成年居民血压水平区域聚集性和危险因素的多水平模型分析 [J]. 中华流行病学杂志, 2014,35(3):246-249.

苏通,李春霞,王君,等. 血液透析患者HBV感染影响因素的多水平模型分析 [J]. 中华流行病学杂志, 2015,36(5):510-514.

吴秋红,张丕德. 广州市某区慢性病患者社区卫生服务利用影响因素研究 [J]. 数理医药学杂志, 2013,26(1):114-116.

Mary JL, Douglas MB. Newton-raphson and EM algorithms for linear mixed-effects models for repeated-measures data [J]. J AM Stat Assoc, 1988,83(404):1014-1022.

金勇进,邵军. 缺失数据的统计处理 [M]. 北京:中国统计出版社, 2009.

Little RJA,Rubin DB. Statistical analysis with missing data(2nd) [M]. New York: John Wiley & Sons, 2002.

Grittner U, Gmel G, Ripatti S, et al. Missing value imputation in longitudinal measures of alcohol consumption [J]. Int J Methods Psychiatr Res, 2011,20(1):50-61.

Schafer JL. Analysis of incomplete multivariate data [M]. London: Chapman & Hall, 1997.

马跃渊,徐勇勇,郭秀娥. MCMC收敛性诊断的方差比法及其应用 [J]. 中国卫生统计, 2004,21(3),154-156.

曹阳,张罗漫. 运用SAS对不完整数据集进行多重填补—SAS 9中的多重填补及其统计分析过程(一) [J]. 中国卫生统计, 2004,21(1):56-63.

Spratt M, Carpenter J, Sterne JA, et al. Strategies for multiple imputation in longitudinal studies [J]. Am J Epidemiol, 2010,172(4):478-487.

Engels JM, Diehr P. Imputation of missing longitudinal data: a comparison of methods [J]. J Clin Epidemiol, 2003,56(10):968-976.

Schafer JL, Olsen MK. Multiple imputation for multivariate missing-data problems: a data analyst's perspective [J]. Multivar Behav Res, 1998,33(4):545-571.

He Y, Yucel R, Raghunathan TE. A functional multiple imputation approach to incomplete longitudinal data [J]. Stat Med, 2011,30(10):1137-1156.

施引文献

资源附件(0)

访问统计