Data and code
The full extracted dataset, the study tables, and the pipeline that produced every number on this site.
The paper
Download the data
The estimate file is the frozen dataset itself: one row per estimate, every coded field, the verbatim quote and the denominator as extracted. It is the file the analysis reads, not an export of it.
The code
Provenance
| Frozen dataset | |
| MD5 | |
| Studies | |
| Estimates | |
| Site built |
Each freeze is immutable, MD5-stamped and git-tagged; the ledger of freezes and the append-only research log ship with the code. Every figure and statistic regenerates from the frozen file by one pipeline, and a drift check fails the build if any document contradicts the canonical counts.
Tell me about a study I missed, or anything else
It reaches a private queue only the author reads; nothing you send is published. If it is a study, a DOI and the quantity as the paper words it is enough, and it is screened against the same eligibility criteria as everything else.