Date-only upper envelope
Definitely after notice and by the selected cutoff.
Loading the frozen local projection...
Visibility envelope
Counts every citation release definitely after a public notice date. It does not prove the warning was visible in a manuscript workflow.
0
indexed citation releases cleared the firewall
Computing from the frozen projection.
0Used: strict post-notice, released by cutoff
0Rejected: later than cutoff
0Rejected: stated date cause
0Undated publication release
0 + 0 + 0 + 0 = 0 complete citation edges
Delay, without invented days
Month- and year-only release dates produce delay intervals. Each bar uses the interval midpoint for drawing; the statistics retain both endpoints.
No points drawn yet.
0Median delay interval
090th percentile interval
099th percentile interval
0Longest possible delay
The complete run
The rule is fixed before later releases are revealed. The replay continues through the entire frozen archive, including reversals in yearly release rate.
All cutoff points are drawn.
| Delay beyond | Definitely | Possibly |
|---|
A warning is not one bit
The notice date proves that a public event preceded a publication release. It does not prove that a usable warning was present when the reference entered the manuscript. The lower envelope asks a narrower historical metadata question: does the current Crossref record contain a Retraction Watch update link, and is its latest deposit timestamp earlier than the citing release interval?
Deposited link lower bound
A Retraction Watch Crossref update link has a deposit timestamp before release.
Marked non-paywalled in 2025
Current snapshot status only. Historical access is not established.
Historical link timing indeterminate
Missing, absent, or not demonstrably deposited before release.
The release ledger
Search targets or citing DOIs. The table lists the longest current delays first and never labels a citation correct or incorrect, because citation context is absent.
| Release | Citing work | Retracted target | Delay interval | Metadata bound |
|---|
No rows shown yet.
The check
Anchor pending.
The pinned 2025-02-28 Retraction Watch CSV is filtered live in the research program through the published 61,645, 18,687, and 16,041 counts.
Firewall pending.
Every citation edge must land in used, later, excluded, or undated. A future-dated plant is inserted into the same runner and must leave the answer unchanged.
Precision pending.
Day, month, year, and missing release dates remain distinct. Same-day and notice-overlapping intervals never enter the strict count.
Apparatus and limits
The cohort bound
All 633 records in the reproduced medical cohort whose Retraction Watch notice date falls in calendar year 2015. Of these, 609 have an original-paper DOI and were queried. The notice year was fixed before citation results were interpreted. It supplies more than ten complete calendar years of follow-up without asking for edges to all 16,041 medical retractions.
The archives
Retraction Watch commit 8c8db609c8d44af184040b56288412fdab41b7d1, released 2025-02-28 23:02 UTC. Citation edges come from the OpenCitations Index API v2, exported 2026-08-14. Its creation field is defined as the citing entity’s publication date. Current link metadata comes from the Crossref REST API. Retraction Watch and OpenCitations data are CC0. The page makes no third-party requests; its projection is local.
Free choices
- Calendar year 2015 is the target notice vintage.
- The complete citation snapshot ends on 2026-08-14.
- For the paper’s unspecified SPSS title-deduplication tie, a same-title record with a non-zero publication-to-retraction interval is retained when available, then the first CSV row.
- A release interval must begin after the notice date to enter the strict count. An interval that touches the notice date is excluded as interval-censored.
- The Crossref lower bound uses the current original-work record’s latest deposit date. Missing evidence stays indeterminate.
What could not have been known
The record does not contain reference-selection date, submission date, acceptance date, historical notice access, editorial screen output, or citation context. No diligence can recover those quantities from publication dates alone. A post-notice release can contain a correct discussion of the retraction. The unobserved branch, what would have happened after a flag, is also unavailable.
Why there is no p-value track
This is an enumeration of a pinned archive, not a sequential claim about a random process. An anytime-valid martingale would add decoration without answering the descriptive question.
Source trail
Pinned Retraction Watch commit, published medical-cohort anchor, Retraction Watch field definitions, and OpenCitations field definitions.