摘要
目的探索研究倾向得分区间匹配法在非随机对照试验中用于均衡组间混杂因素的能力,并与logistic回归分析方法和倾向得分卡钳匹配进行比较。方法通过Monte Carlo模拟分析倾向得分区间匹配法处理二分类资料的能力,并与传统的logistic回归方法以及倾向得分卡钳匹配法进行比较,通过I类错误、检验效能、标准化差异以及匹配比例等指标进行综合评价。结果倾向得分区间匹配法与logistic回归法以及倾向得分卡钳匹配法的检验效能、I类错误、标准化差异和匹配比例四个评价指标无明显差异。结论在观察性研究和流行病学研究中,采用倾向得分区间匹配法均衡组间协变量得到真实的处理效应具有很高的实用价值。
Objective To explore that propensity scores interval matching which was used to balance the confounding factors between groups in non-randomized controlled trials and to compare Logistic regression analysis and propensity score caliper matching.Methods Monte Carlo simulations were applied to estimate the capacity of handling binary data by propensity score interval matching,and to compare type I error rate,test power,standardized difference and matched ratio of Logistic regression,propensity scores caliper matching and propensity scores interval matching.Results There is no significant difference among Propensity score interval matching,Logistic regression and Propensity score caliper matching in Type I error rates,Statistical power,Standardized difference and matched ratio.Conclusion In the observational studies and epidemiological studies,propensity scores interval matching can be used to balance covariates between groups and it is valuable in estimating real treatment effects.
出处
《中国卫生统计》
CSCD
北大核心
2012年第1期53-57,共5页
Chinese Journal of Health Statistics
基金
国家自然科学基金(81001290
81072390)
