Data sources and methodology

The rules behind the numbers. The FAQ has the short answers. This page has the details that matter when you interpret a figure.

Sources

• arXiv is the source of record for titles, abstracts, author lists, categories and versions, read from arXiv's own listings and its OAI-PMH interface and never rewritten.
• INSPIRE-HEP enriches each paper with publication details, citation counts, author identities and affiliations. Papers are matched to INSPIRE by arXiv identifier only.
• Crossref, the DOI registry, supplies publication days where INSPIRE records only the month.

Update schedule

• Digests. arXiv announces new papers Sunday to Thursday at 20:00 US Eastern time. Each listing is fetched during the following hours, so a weekday's digest is normally in place by early morning UTC.
• Citations, journal references, affiliations. Papers from the last twelve months are re-fetched from INSPIRE every night. Older papers are re-checked on a rolling schedule that covers the whole corpus roughly once a month. A citation count on a paper page can therefore be a few weeks old. The page says when it was last observed, and the citation-history chart queries INSPIRE live.
• Journal dates. Where INSPIRE lacks the day of publication, the date comes from Crossref: the online publication date when the journal reports one, otherwise the print date. A "≈" marks the fallback to the DOI's registration day, which can be days early.

Citations

• Counts are INSPIRE-HEP's, so they cover the high-energy and nuclear physics literature rather than all of science. A self-citation, by INSPIRE's definition, is a citation whose citing and cited papers share at least one author. Citations per year divide the count by the time since the arXiv preprint or the journal publication, whichever came first.
• Recent citations. "Citations in the last 90 days" and the What's hot list date each citing paper by INSPIRE's earliest_date for its record: normally the arXiv submission date, otherwise the journal date. When INSPIRE knows only the month or the year, the citation is placed mid-period. This is the same date INSPIRE's own citation timeline uses, so a paper's history here matches its INSPIRE page. The sweep runs nightly, so a citation appears a day or two after INSPIRE indexes the citing paper.
• The cross-list scope never restricts which papers may cite a paper: citation counts and recent-citation figures are the same under either setting.

Cross-list scope

The "Include cross-lists" checkbox next to the corpus name sets the scope for the whole site. Checked, the corpus is every paper listed under the category, as its primary category or as a cross-list. Unchecked, only papers whose primary category is the corpus category count. The digests, Browse, the Authors directory and its statistics, Insights, the Overview and the maps all follow the same rule, so numbers agree across pages. The choice is remembered in a cookie, and a link carrying ?scope=primary shows the primary-only view to anyone. Paper pages always open regardless of scope.

Authors, affiliations and flags

• Identities. Author names come from arXiv and identities come from INSPIRE author profiles. A name is linked to a profile only when the arXiv and INSPIRE author lists of the paper agree position by position, with matching surnames, so that the link is unambiguous. Names that could not be linked stay as plain text without an author page. Names are never merged by spelling alone, so the same person can appear under two names until INSPIRE unifies their profile. Newly created INSPIRE profiles are sometimes provisional and are merged into an existing one later. The nightly refresh picks that up.
• Affiliations. Affiliations are taken from each paper's INSPIRE record. An author's flag and home institution are inferred from their papers: the primary (first-listed) affiliation, weighted toward their most recent papers. The INSPIRE profile's own "current position" is not used, since many profiles keep an outdated one. Both are missing when INSPIRE has recorded no affiliation for the author.
• Map placement. Each institution is placed at the coordinates INSPIRE records for it. Where INSPIRE has none, it is placed at the position of other institutions in the same city, or else at the city itself, using Natural Earth's populated places.

Digest dates

For papers announced since December 2014 the digest date and the position within the mailing are exact, taken from the archived arXiv mailings. For older papers they are reconstructed from the submission time and arXiv's cut-off rules, which agrees with the true announcement day about 91% of the time. Those dates carry an asterisk.

Maps

On the affiliation maps, circle size follows the number of papers with at least one author at the institution. The heat layer shows the density of papers scaled to the current result set, so it is not comparable across different filters.