Less Likely · Bibliometrics
Adoption of the surprisal / S-value transform, live from OpenAlex. Cited = papers citing the canonical methodology. Reported = the subset that actually write the term in their text — a usage signal, not just a name-check.
Metric: deduplicated works citing ≥1 seed paper (a proxy for “used S-values”). The current year is partial. Counts reflect OpenAlex coverage and differ from Scopus/Scholar — trust the trend, not the integer.
Recent papers reporting S-values
Metric: papers that cite a seed paper and mention
"s-value" or surprisal in OpenAlex-indexed full text. This is a
lower bound — full text is indexed for only a subset of papers, and the match
is text-wide, not strictly the Methods section. The strict Methods-only count (Europe PMC open-access
corpus) is ~8, too sparse to chart; the citation∩full-text view here is the cleaner usable signal,
because the citation already fixes the meaning of “S-value.”
marginaleffects (Arel-Bundock, Greifer & Heiss 2024)
prints an S column in its default output, so every citing paper is exposed to an
S-value whether or not it asks for one. This is a fundamentally different signal from the other tabs:
passive exposure, not deliberate use. It is deliberately not merged into the
“Cited” union — only 1 paper cites both marginaleffects and a methodology seed, so the
populations are essentially disjoint. The gap between the two bars is the conversion rate: exposure
rarely turns into reporting. The package’s own manual cites Greenland (2019) and Cole et al.
(“Surprise!”) — two of our three seeds.