did-reviewer
双重差分(DID)实证审查Skill。做DID分析前必须检查平行趋势假设、画图可视化、报告违背情况。触发词:DID审查/双重差分检查/平行趋势/DiD reviewer/difference-in-differences
Install / Use
npx skills add brycewang-stanford/Auto-Empirical-Research-Skills --skill did-reviewerInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
Development & EngineeringSupported Platforms
Our assessment of did-reviewer
did-reviewer scores 88/100 on our quality scale, 782nd of 2,569 Development & Engineering skills we index (top 31%).
Its SKILL.md is 3.3 KB long, well organised into 22 sections with 7 code examples: a solid amount of guidance for an agent.
With 4,360 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated 3 days ago, so did-reviewer is actively maintained.
- No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
- Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
did-reviewer compared with similar skills
All 4 of these similar skills score higher than did-reviewer; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| did-reviewer (this skill)by brycewang-stanford | 88 | 4.4k | 3d ago | SKILL.md |
| Agent-Reachby Panniantong | 100 | 85.6k | 11d ago | CLAUDE.md |
| ai-job-searchby MadsLorentzen | 100 | 44.0k | 6d ago | CLAUDE.md |
| claude-howtoby luongnv89 | 100 | 41.7k | today | CLAUDE.md |
| algorithmic-artby anthropics | 100 | 177.9k | 4d ago | SKILL.md |
Frequently asked questions
- How do I install did-reviewer?
- Run
npx skills add brycewang-stanford/Auto-Empirical-Research-Skills --skill did-reviewer. The install tabs above show the steps for each supported agent. - Which AI agents does did-reviewer work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is did-reviewer safe to use?
- It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is did-reviewer still maintained?
- The repository was last updated 3 days ago, so did-reviewer is actively maintained.
Skill content
View source on GitHubname: did-reviewer description: 双重差分(DID)实证审查Skill。做DID分析前必须检查平行趋势假设、画图可视化、报告违背情况。触发词:DID审查/双重差分检查/平行趋势/DiD reviewer/difference-in-differences version: "1.0" metadata: openclaw: emoji: "📊" homepage: "https://github.com/SiyaoZheng/ai4ss-skills"
did-reviewer — 双重差分 (DID) 实证审查Skill
本 Skill 是 DID(双重差分)实证分析的强制检查清单,在执行 DID 分析前和分析后都必须按本 Skill 检查:平行趋势假设、可视化、违背报告、稳健性。不符合本 Skill 要求的 DID 分析不得发表。
适用场景
- 做 DID 分析前设计检验方案
- 审稿人审查 DID 论文时使用
- 研究者自查 DID 实证结果时使用
- 评估已发表 DID 论文的质量时使用
核心理念
DID 是识别因果的"近似实验",但前提是平行趋势假设成立。若该假设不成立,则 DID 估计量是有偏的,无论结果多么"显著"。
常见错误:
- ❌ 只报告处理组和对照组的系数,不检验平行趋势
- ❌ 平行趋势检验图不加置信区间
- ❌ 不报告政策提前发生的违背情况(anticipation effects)
- ❌ 把 DID 当万能钥匙,忽视识别假设
强制检查清单(DID 七步审查)
✅ 第一步:确认研究设计类型
| 类型 | 命令 | 说明 |
|------|------|------|
| 标准 DID(2×2) | didYw 或手动回归 | 处理组vs对照组,一次政策 |
| 多时点 DID(Staggered DID) | did2/csdid | 不同单位政策时间不同 |
| 交错 DID(Event Study) | eventstudy | 事件前后动态效应 |
| 连续 DID(Continuous DID) | 断点回归变体 | 政策强度连续变化 |
✅ 第二步:平行趋势检验(Pre-trend Test)
必须执行:
- 计算处理组和对照组在政策前期(pre-period)的结果变量均值
- 检验政策前期两组的趋势是否平行
- 使用 事件研究图(Event Study Plot) 可视化
事件研究图规范:
X轴:相对时间(政策前第k期到政策后第k期)
Y轴:处理效应系数
必须包含:置信区间(95%CI)、政策前期系数连线、政策后期系数
判断标准:
- 政策前期(k<0)系数不显著异于0 → 平行趋势成立 ✅
- 政策前期系数显著异于0 → 平行趋势违背 ❌
✅ 第三步:报告违背情况(Violation Report)
若发现以下违背情况,必须如实报告:
| 违背类型 | 描述 | 处理方式 | |----------|------|----------| | Anticipation Effects | 处理组在政策实施前就提前反应 | 将政策前期效应纳入模型 | | Spillover Effects | 对照组受到处理组溢出效应 | 排除溢出污染样本 | | Composition Effects | 处理组样本在政策后发生变化 | 记录并分析样本变化 | | Parallel Trend Violation | 政策前期趋势不平行 | 改用合成控制法(SCM) |
报告格式:
## 平行趋势违背报告
**检验方法**:[描述使用的检验方法]
**违背情况**:[是/否存在违背]
- 若存在:描述违背类型、程度、可能影响
- 若不存在:说明检验结果
✅ 第四步:Staggered DID 特殊检验
若使用多时点 DID / 交错处理,额外检查:
Callaway & Sant'Anna (2021) 检验:
# 使用 csdid 或 did2 包
csdid(lwage, id, time, gvar, control = "nevertreated")
Borusyak et al. (2024) 检验:
# 使用 did 包
did(yname, tname, idname, gname, xformla = ~X)
判断标准:
- 不同处理时间点的效应是否异质
- 是否存在"好坏"处理时间混淆
✅ 第五步:安慰剂检验(Placebo Test)
必须执行:
| 检验类型 | 做法 | 判断标准 | |----------|------|----------| | 时间安慰剂 | 假设政策提前2-3年 | 政策前期系数应不显著 | | 随机处理组 | 随机抽取处理组 | 显著结果应<5% | | 替换变量 | 用同期不相关变量替代因变量 | 应不显著 |
✅ 第六步:稳健性检验
至少报告以下稳健性检验:
- 排除异常值:剔除处理组中政策前后的极端值
- 不同控制组:替换对照组的定义
- 不同模型设定:增减控制变量、改变函数形式
- 样本敏感性:不同样本范围的结果一致性
✅ 第七步:异质性分析(可选但建议)
报告处理效应的异质性:
| 异质性维度 | 分组方式 | |------------|----------| | 地区 | 沿海vs内陆 | | 企业规模 | 大vs小 | | 行业 | 高技术vs低技术 | | 时间 | 政策早期vs晚期 |
输出标准格式
## DID 实证审查报告
**审查对象**:论文标题/研究问题
**使用的识别策略**:标准DID / 交错DID / 连续DID
**数据来源**:[数据库名称,附provenance]
### 审查结果
| 检查项 | 状态 | 说明 |
|--------|------|------|
| 平行趋势检验 | ✅/❌ | 检验方法和结果 |
| 事件研究图 | ✅/❌ | 是否规范 |
| 违背情况报告 | ✅/❌ | 是否如实报告 |
| 交错DID特殊检验 | ✅/❌ | 使用的方法 |
| 安慰剂检验 | ✅/❌ | 结果是否合理 |
| 稳健性检验 | ✅/❌ | 报告了哪些 |
| 异质性分析 | ✅/❌ | 分析内容 |
### 致命缺陷(如有)
[列出可能导致识别策略失效的致命问题]
### 建议修改
[针对性的修改建议]
调用接口
在 Claude Code 对话窗口输入:
/did-reviewer
或完整 Prompt:
按DID审查Skill检查本项目的DID分析:必须包含平行趋势检验、事件研究图、违背情况报告、交错DID特殊检验、安慰剂检验、稳健性检验。输出结构化审查报告。
与其他 Skill 的配合
grillme→ 选题时确定使用 DID 作为识别策略citation-fidelity→ 引用 DID 相关方法论文献时核查准确性econ-reviewer→ 完整论文审查时配合使用
重要声明
- 本 Skill 的审查标准参照**顶刊(AER, Econometrica, QJE, JPE)**要求
- 平行趋势检验是必要条件,不是可选项
- 发现违背情况不等于论文无效,但必须如实报告,审稿人会判断是否接受
Related Skills
Agent-Reach
85.6kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
ai-job-search
44.0kThe job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.
claude-howto
41.7kA visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates that bring immediate value.
algorithmic-art
177.9kCreating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
