Cleaning steps

[nniiicc]

@sarah114tran - documenting data collection and some preliminary cleaning of the dataset here.

Two new directories in the repo:

• Data - contains all data collected.
• Code - Scripts used to clean the data

Next steps:

• Combine the dataset titled dedup_clean.csvwith our previously coded dataset from ASReview (where we agreed on what to include and not include)
• Reingest to ASReview
• Begin coding again with our strict criteria on what counts for empirical:
• • Report results about OSH (and not OSS for example)
• • Uses method that is repeatable or quasi-scientific (e.g. case study, survey, statistical analysis, etc)

---

Query:

• ("open-source hardware" OR "open source hardware" OR "open hardware")
• Web of Science - 774 (1000)
• Crossref 1000 (1000)
• Scopus 200 (200)
• Semantic Scholar 1000 (1000)
• Google Scholar 370 (500)*

For Google Scholar query:

• ("open-source hardware" OR "open source hardware" OR "open hardware") AND "empirical"

Total of 3474 observations

Applying inclusions and exclusion criteria

Date Removed out of range - 661

Deduplication: Original number of rows: 2813 Number of duplicates identified by title: 959 Number of rows after title de-duplication: 2241 Number of duplicates identified by DOI: 473 Final number of rows after DOI de-duplication: 1802

Crossref does not allow for search by type... so clean by type (removed following)

• 5 - PDF notes
• 20 REVIEWS
• 6 REPORTS
• 58 pre-prints
• 3 abstracts
• 1 grant
• 7 support materials
• 7 datasets

Final dataset: 1692

[sarah114tran]

@nniiicc -- I have decided to drop all of the articles with empty abstracts, because it can be hard to tell with just title will document in the coding notes/protocol I would drop the null values in the in the "abstract" column in the dedup_clean.csv before combining it with your previously coded dataset, because the capitalization of each column label in not exact in both datasets.