METHOD · mth_psm · v2
Propensity Score Adjustment
倾向得分调整
Uses treatment propensity conditional on observed covariates to balance measured confounders via matching, weighting, stratification or adjustment.
方法身份
- 规范名称
- Propensity Score Adjustment
- 别名
- propensity score、PSM、propensity score matching
- 方法族
- observational causal adjustment
以上名称和分类来自此版本资料卡,不代表方法之间存在已核验的关系或适用性。
从资料卡到研究设计
编辑整理的设计输入与可能输出;并非平台支持的运行接口、实际分析结果或专家适用性核验。
编辑整理日期:2026-10-10 · 专家适用性尚未核验
核对输入与输出的来源 ↗需要准备的输入
- 处理指标、处理前协变量、结果及目标人群
- 匹配或加权规则、重叠范围与目标量
可能产生的输出
- 匹配样本、权重、平衡与重叠诊断
- 效应需要后续适当分析;平衡不排除未测量混杂
适用问题
- observational treatment comparisons with rich pre-treatment covariates
数据条件
- treatment
- pre-treatment covariates
- outcome
- overlap
识别与模型假设
- conditional exchangeability on measured covariates
- positivity/overlap
- valid adjustment procedure
常见失败
- post-treatment covariates
- prediction instead of balance
- assuming measured balance removes unmeasured confounding
不能支持的结论
- control of unmeasured confounding
- causality under severe non-overlap
诊断与核对
- post-adjustment balance
- overlap
- weight diagnostics
- sensitivity
可比较的替代方法
- doubly robust estimation
- stronger quasi-experimental designs
软件与实现
- R MatchIt/WeightIt
- Stata teffects
来源与参考
- The central role of the propensity score in observational studies for causal effects
Paul R. Rosenbaum; Donald B. Rubin · 1983
打开原始来源 ↗
来源链接用于核对。来源可访问不代表已获全文或代码再分发授权。
对象关系
关系连接明确的对象版本。前置知识指学习或实现知识;搭配关系是编辑建议,尚未统计真实使用频率。
Corporate Governance and Capital Markets →
Measured-confounding adjustment when assumptions fit
版本、依据与适用边界
top_corporate_governance v2 → mth_psm v2
editorial mapping grounded in method scope; domain-editor review required; snapshot bound to the explicitly selected object revisions
核对关系来源 ↗关系 kre_tme_021 · v1 · 待领域复核