METHOD · mth_rdd · v2
Regression Discontinuity Design
断点回归
Identifies a local effect at a cutoff when assignment changes discontinuously with a running variable and potential outcomes are suitably smooth absent treatment.
方法身份
- 规范名称
- Regression Discontinuity Design
- 别名
- RDD、RD design
- 方法族
- causal inference / threshold design
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
- 运行变量、阈值、处理状态与结果变量
- 核函数、带宽与阈值附近样本规则
可能产生的输出
- 阈值附近的局部效应估计与区间
- 带宽敏感性和阈值核对;不直接推广为全人群平均效应
适用问题
- eligibility/treatment triggered by a threshold
数据条件
- running variable
- known cutoff
- observations near cutoff
识别与模型假设
- no precise manipulation in canonical design
- continuity/smoothness of potential outcomes
- correct assignment mechanism
常见失败
- high-order global polynomial
- arbitrary bandwidth
- extrapolating local effect
不能支持的结论
- population-wide ATE without extra structure
- effect if another policy also changes at cutoff
诊断与核对
- density/manipulation
- covariate continuity
- bandwidth sensitivity
- local polynomial/robust inference
可比较的替代方法
- DiD
- IV
- matched observational designs
软件与实现
- R rdrobust
- Stata rdrobust
来源与参考
- Regression discontinuity designs: A guide to practice
Guido W. Imbens; Thomas Lemieux · 2008
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。
Policy Evaluation →
Threshold-based assignment
版本、依据与适用边界
top_policy_evaluation v2 → mth_rdd 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_004 · v1 · 待领域复核
Education Policy and Inequality →
Eligibility/admission thresholds
版本、依据与适用边界
top_education_inequality v2 → mth_rdd 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_009 · v1 · 待领域复核