Research Code Quality and Execution Analysis for 2000+ Datasets, 2010-2020

More than 2000 replication datasets with over 9000 unique R files from the Harvard Dataverse repository, published between 2010 and 2020. The dataset accompanies a study by Ana Trisovic, Matthew K. Lau, Thomas Pasquier, and Mercè Crosas that analyzes code quality and execution success. The study found 74% of R files failed to execute without error initially, with failures reduced to 56% after automatic code cleaning.

Use Cases

• Analyzing common coding errors in research code based on the execution failure rates mentioned in the description
• Studying the impact of journal policy strictness on code re-execution rates as referenced in the abstract
• Developing automated code cleaning tools based on the error patterns identified in the large-scale study
• Formulating recommendations for code dissemination aimed at researchers, journals, and repositories as described in the article

Strengths

• Analysis of over 2000 replication datasets, providing a large-scale view of research code practices
• Examines more than 9000 unique R files, offering substantial sample size for code quality assessment
• Covers a 10-year time range (2010-2020), allowing for longitudinal analysis of trends

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Last update date is unknown; freshness unverified

Provenance

Source

Harvard Dataverse repository

Collection Method

Retrieved and analyzed replication datasets published with academic papers

Time Range

2010 to 2020