Replication package (Level 2, restricted) for "Auditing successor selection in software repository mining: Evidence from filtering AI-attributed commits"

Abstract

Level-2 (restricted) archive of the replication package of the article Auditing successor selection in software repository mining: Evidence from filtering AI-attributed commits (H. Jeong, Hanyang University ERICA; manuscript submitted to the Journal of Systems and Software). The public Level-1 archive with the programs, tests and aggregate results is https://doi.org/10.5281/zenodo.23226959. The files are per-unit derived counts of the 3,687 cohort-A accounts and the cohort-B identities, with random identifiers (marker dates reduced to the month; no account names, e-mail addresses, repository names, commit SHAs or commit times). Placed by python3 reproduce.py --level2 of the Level-1 package, they recompute the registered aggregates and the exploratory aggregates that do not require the raw commit logs, including rates, transitions, paired contrasts, random-exclusion benchmarks and bootstrap intervals; on Python 3.10.12 with numpy 2.2.6 every regenerated result agreed with the reported one (numerical values within 1e-9, text files byte-identical). Access is restricted because cohort-B rows keep the positions of their repositories in the public sampling order, which the registered repository-cluster bootstrap needs. Access is granted to researchers who state their name, affiliation and purpose and who agree to use the files only to verify, reproduce or extend the analyses of the article, not to redistribute the files or per-unit derivatives, not to attempt any re-identification or linkage to public commit histories or accounts, to store the files securely and delete them when the work is complete, and to cite the article and both archives (ACCESS.md). Reviewers of the article receive an access link through the editor.