From Search Triggers to Catalog Events

A search trigger is a pipeline-specific excursion of a ranking statistic. A catalog event is a versioned scientific record assembled from associated triggers, detector and data-quality context, false-alarm information, classification assumptions, and—only for a selected subset—parameter-estimation products. Treating every trigger as an event, or every event as statistically interchangeable, breaks the chain from detector data to catalog inference.

Helpful background. Coincident and Coherent Network Searches explains how pipelines form network candidates; False-Alarm Rates and Trials defines the background quantities used here.

Four distinct objects

The following terms answer different conditional questions:

Object	Minimal content	It is not yet
Trigger	Pipeline, detector/time, template or morphology, ranking statistic	A unique astrophysical source
Candidate	Associated network triggers with data-quality context	A catalog entry under every threshold
Catalog event	Versioned record satisfying declared inclusion and association rules	A population-independent fact
Parameter-estimation product	Posterior samples and analysis metadata for a chosen event/version	A likelihood or population posterior by itself

A trigger’s ranking statistic may combine signal consistency, detector sensitivity, coincidence, template class, and background information. Its numerical value has meaning only within that pipeline and analysis version. Different pipelines need not rank the same data identically.

Association and de-duplication

Suppose triggers $a$ and $b$ come from different searches. An association rule might require compatible geocentric time, source class, and template or morphology:

$$\lvert t_a-t_b\rvert<\Delta t, \qquad D(\vartheta_a,\vartheta_b)<\Delta_\vartheta.$$

These conditions define edges in a trigger-association graph. Connected components are candidate groups, but transitivity can produce a chain whose endpoints are not mutually compatible. A production rule therefore also needs deterministic tie-breaking, conflict resolution, and a representative-record choice.

De-duplication must preserve provenance. Combining two records should not discard their pipeline-specific ranking, background, detector set, data-quality flags, or source-class assumptions. Those differences are later needed to reconstruct thresholds and selection functions.

Data quality and analysis readiness

Catalog construction conditions on analyzed detector time, not wall-clock observing time. A candidate record should identify:

• participating detectors and the calibrated data release;
• the exact analysis segment and category of data-quality selections;
• gating, subtraction, or other transient treatment;
• the search and pipeline version;
• the ranking statistic and background estimate;
• the association/de-duplication rule; and
• the catalog release in which the record appears.

Data-quality information can veto an instrumental artifact or change the valid background. It should not be retrofitted as an unexplained label after a threshold has been applied. The detector-noise guide describes the connection between glitches, data quality, searches, and parameter estimation LIGO and Virgo 2020, §§ 3–5.

From candidate to event-level products

Catalog inclusion and parameter estimation are separate selections. A catalog can contain candidates without full source-parameter products, while parameter estimation may use a different data segment, waveform family, calibration product, or frequency band from the search. The event record must connect, not conflate, these stages:

$$\text{strain version} \longrightarrow \text{pipeline trigger} \longrightarrow \text{associated candidate} \longrightarrow \text{catalog decision} \longrightarrow \text{event analysis version}.$$

False-alarm rate and astrophysical probability are not interchangeable. The former is a property of a search background and threshold; the latter additionally assumes signal and noise populations. Parameter posterior samples are conditional on event-level priors and likelihood choices, not on the search ranking statistic alone.

A reproducible toy construction

A small independent test can use a machine-readable trigger table with fields

$$(\text{pipeline},t,\text{class},x,\mathrm{FAR},\text{detectors},\text{DQ version}),$$

where $x$ is the pipeline ranking statistic. The procedure is:

1. freeze the input table and a deterministic association rule;
2. group compatible triggers without losing their original rows;
3. apply pipeline-specific thresholds;
4. choose a representative candidate by a declared rule;
5. attach event-analysis products only when their exact version exists; and
6. serialize the catalog plus provenance and a content hash.

Vary the association window, threshold, and tie-breaker. The output should change in predictable rows while untouched records remain byte-for-byte stable. Failure modes include chain associations, duplicated sources, a candidate surviving only because backgrounds from incompatible pipelines were compared directly, and stale parameter products attached to a revised event.

To make the toy construction reproducible, retain the trigger table, code and dependency versions, input hash, random seed if used, and expected row-level differences. This tests association logic; reproducing a collaboration catalog additionally requires its real search, background, and data-quality inputs.

Current release boundary

The empirical boundary is 2026-08-18. The exact current source is the GWTC-5.0 data release, with the search, candidate, sensitivity, and event-analysis construction described in GWTC-5.0 Methods, §§ 2–5 and its catalog scope reported in GWTC-5.0 Observations, §§ 2–4 and release tables. No event count is made evergreen here. A later minor release, corrected product, changed association, or withdrawn candidate creates a new evidence state rather than silently rewriting this one.

Exercise

A toy table contains triggers

Label	Pipeline	Time (s)	Class	FAR ($\mathrm{yr}^{-1}$)
A	P	100.00	CBC	0.02
B	Q	100.06	CBC	0.05
C	P	100.18	CBC	0.01
D	Q	240.00	burst	0.03

Associate triggers of the same class when their times differ by at most $0.10\,\mathrm{s}$. Find the naive connected components and identify the transitivity hazard.

Solution

A is associated with B because their separation is $0.06\,\mathrm{s}$. B is not associated with C: the separation is $0.12\,\mathrm{s}$, and A is also not associated with C. Thus the components are $\{A,B\}$, $\{C\}$, and $\{D\}$; D also differs in class.

The hazard appears if B were at $100.09\,\mathrm{s}$ and C at $100.18\,\mathrm{s}$. Then A–B and B–C would both be edges, so a connected-component rule would group all three even though A–C are $0.18\,\mathrm{s}$ apart. A production association needs a representative-time or all-pairs compatibility rule, not unexamined transitive closure.

References

• LIGO Scientific Collaboration and Virgo Collaboration (2020). “A Guide to LIGO–Virgo Detector Noise and Extraction of Transient Gravitational-Wave Signals.” Classical and Quantum Gravity 37, 055002. DOI.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration (2026). “GWTC-5.0: Methods for Identifying and Characterizing Gravitational-Wave Transients.” arXiv:2605.27224. Open PDF.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration (2026). “GWTC-5.0: Observations from the Second Part of the Fourth LIGO-Virgo-KAGRA Observing Run and Updates to the Gravitational-Wave Transient Catalog.” arXiv:2605.27225. Open PDF.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration (2026). GWTC-5.0 Data Release. DOI 10.7935/bk00-6a89. Official release.