Document Type
Abstract
Publication Date
2026
Conference Name
4th International Conference on the Science of Science & Innovation
Abstract
Reproducibility and replicability (R&R) are structural properties of scientific knowledge, yet existing R&R evidence remains fragmented across domains and initiatives. We present an ongoing effort to develop a centralized, cross-domain database of R&R studies that links published works to their corresponding R&R attempts and supporting assessments. Bibliographic metadata and title-based matching through scholarly indexing services identify canonical records and persistent identifiers. A heuristic, uncertainty-aware matching algorithm supports intra-source and inter-source deduplication, complemented by manual review of ambiguous cases. A unified schema accommodates heterogeneous assessment frameworks and records original studies, R&R studies, assessments, aggregated summaries, and source provenance. The database currently contains 566 original studies and 616 R&R studies from five sources: MRR, CORE-Bench, ReScience C, MLRC, and I4R. R&R studies span Health Sciences (5.68%), Life Sciences (11.53%), Social Sciences (47.24%), and Physical Sciences (35.55%). By consolidating scattered evidence into a unified, queryable resource, the database supports systematic estimation of R&R rates, investigation of potential predictors, and comparison of reproducibility patterns across fields and over time. Active ingestion is ongoing, and a public web interface for browsing, querying, and visualizing dataset snapshots is under development.
Repository Citation
Obadage, Rochana R.; Rajtmajer, Sarah; and Wu, Jian, "Toward a Centralized Cross Domain Database for Reproducibility and Replicability Studies" (2026). Computer Science Faculty Publications. 479.
https://digitalcommons.odu.edu/computerscience_fac_pubs/479
ORCID
0000-0003-1593-4052
Included in
Databases and Information Systems Commons, Data Science Commons, Scholarly Communication Commons