SMR Software and Reproducibility
Use this to meet SMR's defining expectation: a methods paper should ship a method other people can
run. SMR readers adopt methods, so released, documented software and a reproducible pipeline are
not optional extras — a beautiful estimator with no usable code is a method no one will use, and
reviewers know it.
What "released software" means at SMR
- A package or well-structured code implementing the method in a mainstream environment (R, Stata,
Python — whatever the audience uses), with a documented interface, not a private analysis script.
-
Worked usage: a minimal example that runs the method on toy or example data, so a reader can
confirm it works before adapting it.
-
Versioned and citable: deposit in a trusted public repository (e.g., a tagged release with a DOI)
rather than a personal homepage that can disappear.
The reproducibility pipeline
Every exhibit in the paper must be regenerable:
- A README stating software versions, dependencies, expected runtime, and the entry-point command.
- A master script that runs the full pipeline: simulation grid → tables → figures → illustration,
with seeds fixed so Monte Carlo results reproduce exactly.
-
One command per exhibit mapping (or a clear table) so a replicator knows which script produces
Table 2, Figure 3, etc.
- For long simulations, provide a smoke-test mode (few replications) that runs quickly and a
full-run mode that reproduces the paper.
The SMR availability statement
SMR requires a data-and-code availability statement (检索于 2026-06;以官网为准):
- Disclose how the study data, code, and materials can be permanently accessed — repository name,
DOI/URL, and any access conditions.
-
Code and materials are expected; data should be deposited where possible, and restricted data
must state the reason and the access path/conditions in the statement.
- Use a trusted public repository; reference datasets in DataCite style.
- Write the statement at submission, not at acceptance — it is part of the manuscript's conformance.
Anonymization of the package under double-anonymized review
- The released materials submitted for review must not deanonymize: strip author names, institutional
paths, and account handles from code, README, and repository metadata.
- Use an anonymized repository (e.g., a blinded archive) for the review copy; reveal the public,
citable version at acceptance.
Checklist
Anti-patterns
-
Private analysis script as "software": code that only the authors can run.
-
Unseeded simulation: Monte Carlo results that cannot be reproduced.
-
Disappearing host: code on a personal homepage with no archive/DOI.
-
Missing exhibit map: a replicator cannot tell which script makes which table.
-
Availability statement deferred to acceptance, or vague ("available on request").
-
Deanonymizing repo metadata in the review copy.
Output format
[Reproducibility status] release-ready / needs work / not ready
[Software] package/structured code? usage example? citable deposit?
[Pipeline] master script + seeds + exhibit map present? smoke-test mode?
[Availability statement] written? repository/DOI + access conditions stated?
[Anonymization] review-copy materials clean? yes/no
[Next SMR skill] smr-submission