jbes-replication-and-data-policy - SKILL.md Agent Skill

name: jbes-replication-and-data-policy description: Use when assembling the reproducible data-and-code supplement for a Journal of Business & Economic Statistics (JBES) paper under the American Statistical Association (ASA) data-sharing and reproducibility policy. Builds the supplement; it does not run the analysis itself.

Replication & Data Policy (jbes-replication-and-data-policy)

When to trigger

The paper is heading toward submission/acceptance and a code+data supplement is needed
You want a reproducible pipeline early to pre-empt referee replication requests
Raw data are restricted/proprietary and you need a disclosure-compliant plan
You need to write a data availability statement

Why this matters at JBES

JBES is owned by the American Statistical Association and published by Taylor & Francis, so it follows ASA's data-sharing and reproducibility policy rather than the AEA Data Editor regime. Under that policy, the ASA strongly encourages authors to include data sets and code as supplementary material that demonstrate the article's results, to make the underlying data publicly available, and to include a data availability statement; individual ASA journal editors may set their own stricter rules, and authors may request a waiver for confidentiality/security reasons. Important: JBES does NOT use the JAE Data Archive — that archive belongs to the Journal of Applied Econometrics, a separate Wiley journal. Treat any stricter JBES-specific enforcement as a live T&F instructions check.

Supplement anatomy (methods paper)

supplement/
  README.{md,pdf}      # data sources, software versions, run order, runtime, exhibit map
  code/
    00_setup           # installs/pins packages, sets paths, sets seeds
    01_simulation      # regenerates every Monte Carlo table/figure
    02_application     # regenerates every empirical table/figure
  data/
    raw/               # as obtained, or access instructions if restricted
    derived/           # built by code, never hand-edited
  output/              # regenerated, checked against the manuscript

README must specify

Provenance of every data source (version/vintage, access date); for restricted data, exact access steps.
Software environment with pinned versions (renv.lock, requirements.txt, recorded ssc/net + Stata version).
Run order via one master script; expected wall-clock runtime, hardware, and total Monte Carlo replications.
Seeds for every simulation/bootstrap/randomization step.
Exhibit map linking each table/figure (including simulation tables) to the script that produces it.

Checklist

One master script regenerates every table and figure, including all Monte Carlo numbers
Software and package versions pinned and documented
Random seeds set and reported for all stochastic steps
Data availability statement drafted (ASA expectation)
Restricted/proprietary data have documented access steps + a runnable proxy
No hand-edited derived data; absolute paths removed
Output regenerated and checked against the manuscript numbers
Live JBES-specific enforcement checked if submission-ready advice is being given

Anti-patterns

A zip of scripts with no README and no run order
Hard-coded local paths that break on another machine
Missing seeds, so Monte Carlo numbers cannot be reproduced
Assuming the JAE Data Archive applies (it does not — that is a different journal)
Asserting JBES mandates deposit beyond ASA policy without checking the live T&F instructions

Worked vignette: building a supplement for a Bayesian VAR forecasting paper

A hypothetical JBES paper proposes a hierarchical-shrinkage Bayesian VAR and forecasts US macro aggregates from FRED-MD (details illustrative). The supplement README pins R 4.3 with renv.lock, records the FRED-MD vintage (2024-03) and access date, and sets the Gibbs-sampler seed plus an illustrative 20,000 draws after a 5,000 burn-in. One master script runs 01_simulation (coverage tables) then 02_application (forecast figures), with an exhibit map tying Table 3 to its script and a stated 40-minute wall-clock on documented hardware. Because every posterior summary is seed-pinned, a referee who reruns it reproduces the numbers exactly — pre-empting the "I could not regenerate Table 3" report.

Reproducibility-pushback patterns (venue-specific fixes)

JBES referee/editor objection	Fix this skill enforces
"I could not reproduce the Monte Carlo numbers."	Set and report seeds for every stochastic step; pin the replication count
"Your scripts assume your machine."	Remove absolute paths; pin package versions via `renv.lock`/`requirements.txt`
"Where is the data availability statement?"	Draft it per ASA expectation; for restricted data give access steps plus a runnable proxy

Calibration anchor (hedged): JBES follows ASA data-sharing and reproducibility policy under Taylor & Francis — not the AEA Data Editor regime and not the JAE Data Archive (a different journal). ASA strongly encourages a code-and-data supplement and a data availability statement; check live T&F instructions before asserting stricter JBES-specific enforcement.

Output format

【Master script】regenerates all tables/figures incl. Monte Carlo? [Y/N]
【Env pinned】versions + packages documented? [Y/N]
【Seeds】set + reported? [Y/N]
【Data availability statement】drafted? [Y/N]
【Restricted data】access steps + runnable proxy? [Y/N / NA]
【Policy check】live JBES-specific enforcement checked if needed? [Y/N]
【Next step】jbes-review-process