Background/aims: While randomized controlled trials are the gold standard for measuring causal effects, robust conclusions about causal relationships can be obtained using data from observational studies if proper statistical techniques are used to account for the imbalance of pretreatment confounders across groups. Propensity score (PS) and balance weighting are useful techniques that aim to reduce the observed imbalances between treatment groups by weighting the groups to be as similar as possible with respect to observed confounders. Methods: We have created CoBWeb, a free and easy-to-use web application for the estimation of causal treatment effects from observational data, using PS and balancing weights to control for confounding bias. CoBWeb uses multiple algorithms to estimate the PS and balancing weights, to allow for more flexible relations between the treatment indicator and the observed confounders (as different algorithms make different (or no) assumptions about the structural relationship between the treatment covariate and the confounders). The optimal algorithm can be chosen by selecting the one that achieves the best trade-off between balance and effective sample size. Results: CoBWeb follows all the key steps required for robust estimation of the causal treatment effect from observational study data and includes sensitivity analysis of the potential impact of unobserved confounders. We illustrate the practical use of the app using a dataset derived from a study of an intervention for adolescents with substance use disorder, which is available for users within the app environment. Conclusion: CoBWeb is intended to enable non-specialists to understand and apply all the key steps required to perform robust estimation of causal treatment effects using observational data.
翻译:背景/目标:尽管随机控制试验是衡量因果关系的黄金标准,但如果使用适当的统计技术来计算各组预处理混乱者之间的不平衡情况,则可利用观测研究的数据,得出关于因果关系的可靠结论。 分数分数(PS)和平衡加权是有用的技术,目的是通过加权各组对治疗群体之间观察到的不平衡情况进行减少所观察到的不平衡,使其与观察到的混结者尽可能相似。方法:我们已经创建了CoBWeb,这是一个自由、易于使用的网络应用程序,用于估算观测数据造成的因果处理效果,使用PS和平衡权重来控制因果关系。CoBWeb使用多种算法来估计PS和权衡权重之间的不平衡。CobWeb使用多种算法来估算治疗指标与观察到的共体(不同的算法使得对治疗变异体和混集体之间的结构关系进行尽可能灵活的假设。 最佳算法可以通过选择一种从观测数据中实现平衡和有效抽样大小之间最佳交易的网络处理方法。 结果:CoWeb遵循多种关键的方法来估计方法来估计PS值的准确性观测效果,我们从进行可靠的数据分析,然后从进行可靠的数据分析,以便进行可靠的数据分析。