Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› expecon-replication-package
expecon-replication-package
GitHub用于组装实验经济学论文的数据、代码和说明，以符合ESA可重复性标准。构建提交包，包含指令、软件、原始数据及分析脚本，确保结果可复现与实验可运行。
Trigger Scenarios
准备提交实验经济学论文并需附上参与者说明和数据附录时
尚未组装符合ESA政策要求的可重复性材料存档时
审稿人或编辑要求确认能否复现数据或重新运行实验时
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill expecon-replication-package -g -y
SKILL.md
Frontmatter
{
"name": "expecon-replication-package",
"description": "Use when assembling the data, code, instructions, and experiment software for an Experimental Economics (ExpEcon) manuscript to meet the ESA reproducibility standard. Builds the deposit; it does not run the analysis or draft prose."
}
Replication Package (expecon-replication-package)
When to trigger
- You are preparing to submit and must attach participant instructions (required at submission) and a data/code appendix
- The ESA Data and Replication Policy deposit (trusted repository) is not yet assembled
- z-Tree / oTree code, raw session data, and analysis scripts are scattered and not runnable end-to-end
- A referee or editor asks whether someone could reproduce your numbers and re-run your experiment
What ExpEcon reproducibility actually requires
Experimental Economics is an ESA journal, and since 2021 the ESA Data and Replication Policy requires authors to deposit, in a trusted online repository, the materials needed to reproduce or replicate the study (检索于 2026-06；以官网为准). Reproducibility here is stronger than at most economics journals because it has two layers:
- Reproduce the analysis — raw data + cleaning + analysis code regenerate every table and figure.
- Replicate the experiment — instructions + experiment software let another lab re-run the study.
Treat the package as a deliverable engineered for both.
The deposit, component by component
- Instructions — the exact instructions subjects received, per treatment, in the original language (translation if relevant). These are required at submission, not just at acceptance; reviewers read them to check for deception and comprehension.
- Experiment software — the z-Tree
.ztttreatment files or the oTree app (full project,settings.py, requirements pinned). Include screenshots or the comprehension quiz as run. This is what makes re-running possible. - Raw data — session-level exports as collected (z-Tree
.xls/.sbj, oTree CSV), with a codebook for every variable and the session/treatment/matching-group identifiers. - Analysis code — scripts (Stata/R/Python) that run from raw to results with a single master file; set and record the random seed for any simulation/permutation test.
- README — repository map, software versions, run order, expected runtime, and a table mapping each exhibit in the paper to the script that produces it.
- Pre-registration / PAP link — the registry entry and timestamp; for a Registered Report, the in-principle-acceptance Stage-1 protocol.
- Ethics / consent — IRB approval reference and the consent procedure (and the explicit no-deception statement).
Repository and hygiene
- Deposit in a trusted, persistent repository (OSF, Harvard Dataverse, Zenodo, or OpenICPSR are commonly used by ESA authors) and cite the DOI in the paper.
- Anonymize subject identifiers; never include payment records with identifying info.
- Pin every dependency and software version; a package that does not run on a clean machine fails the policy.
- Match repository contents to the paper exactly — no stale scripts, no figures the code cannot produce.
A workable directory layout
/instructions treatment_A.pdf, treatment_B.pdf (+ translations)
/software ztree/ *.ztt OR otree/ (full app, requirements.txt)
/data/raw session exports as collected (.xls/.sbj or .csv)
/data/clean analysis-ready files built by /code
/code 00_master.* , 01_clean.* , 02_analysis.* , 03_figures.*
/output tables + figures regenerated by /code
README.md map, versions, run order, exhibit→script table
ETHICS.md IRB ref, consent text, no-deception statement
The single rule the policy enforces in spirit: a stranger with a clean machine runs 00_master and gets your paper's exact numbers, and another lab opens /software and /instructions and re-runs your experiment.
The two-layer self-test
- Reproduce: delete
/data/cleanand/output, run the master script, confirm every table/figure regenerates byte-for-byte (or value-for-value for stochastic steps with a fixed seed). - Replicate: hand
/software+/instructionsto a colleague who was not on the project and confirm they can launch a session and understand what subjects faced.
Checklist
- Participant instructions (all treatments, original language) included at submission
- z-Tree
.ztt/ oTree app deposited so the experiment can be re-run - Raw session data + codebook + session/group/treatment IDs present
- Master analysis script runs raw→results; seeds set for simulation/permutation
- README maps every table/figure to the script that generates it; versions pinned
- Pre-registration / PAP (or Stage-1 RR protocol) linked with timestamp
- Trusted-repository DOI cited; data anonymized; IRB + no-deception statement included
Anti-patterns
- Promising the package "on request" or only at acceptance — ESA expects a real deposit, and instructions are due at submission
- Depositing data but not the z-Tree/oTree code, so the experiment cannot be replicated
- A "replication package" whose scripts do not reproduce the paper's exact numbers
- Unpinned software versions / no seed, so permutation tests and figures are not reproducible
- Identifiable subject data or payment records left in the repository
Output format
【Journal】Experimental Economics (ESA method flagship)
【Skill】expecon-replication-package
【Verdict】deposit-ready / incomplete
【Instructions】all treatments, at submission? [Y/N]
【Software】z-Tree .ztt / oTree app deposited (re-runnable)? [Y/N]
【Data + code】raw + codebook + master script (seeded) reproduce all exhibits? [Y/N]
【Repository】trusted-repo DOI; versions pinned; anonymized? [Y/N]
【Pre-reg / ethics】PAP/RR link + IRB + no-deception statement
【Next skill】expecon-referee-strategy
Version History
- 1839142 Current 2026-07-05 13:13
