【作者】 张亮； 李婵娟； 夏结来；

【机构】 第四军医大学卫生统计学教研室；

【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。更多还原

【Abstract】 objective:To explore the ability of propensity scores interval matching balances the confounding factors between the groups in non-randomized controlled trials,with the Logistic regression analysis and propensity score caliper matching for comparison.Methods:Monte Carlo simulations were applied to score the capacity of handling binary data by propensity score interval matching,then compare Logistic regression,propensity scores caliper matching and propensity scores interval matching by type I error rate, test power,standardized difference and matched ratio.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 the groups.The method has a high practical value to estimate the real treatment effects.更多还原