Glossary

background

The accidental coincidence rate—how often noise fluctuations produce triggers at a given ranking statistic. Estimated from non-zero lag data where real signals are broken by time-shifting one detector.

burst

A short-duration gravitational-wave signal without a precise waveform model. Sources include supernovae, gamma-ray bursts, and neutron star mergers.

CBC-PE

Compact-binary-coalescence parameter estimation. In public GWTC comparisons, cWB burst reconstructions are often compared with waveform-model-based CBC parameter-estimation results.

cc

Network correlation — normalized cross-correlation between detectors measuring how well data matches the expected signal model. Threshold: netCC (default 0.5).

CED

Coherent Event Display — a cWB event report page with plots, reconstructed waveforms, sky information, and diagnostic views for a candidate or public event.

chi2

(\(\chi^2\)) — goodness-of-fit statistic. \(\chi^2 \sim 1\) for well-modeled signals, \(\chi^2 > 1\) for glitches.

cluster

A group of time-frequency pixels with excess coherent power, selected before likelihood evaluation. Multiple clusters are merged into superclusters.

coherent

Appearing consistently across all detectors after accounting for time-of-flight delays. Real GW signals are coherent; instrumental noise generally is not.

config

Short for user_parameters.yaml — the single YAML file controlling all aspects of a pycWB search.

cWB

Coherent Wave Burst — the core algorithm that pycWB implements. Uses wavelet transforms and network coherence to detect unmodeled bursts.

cWB-2G

The standard second-generation coherent WaveBurst pipeline used for many LVK burst analyses and public GWTC waveform reconstructions.

cWB-XP

A separate cWB branch described in the legacy public pages. It keeps the ROOT/C++ framework but uses a WaveScan-style multi-resolution transform and a cross-power statistic rather than being part of pycWB.

DPF

Dominant Polarization Frame — the optimal polarization basis for a given sky direction. Maximizes the signal projected onto the \(+\) polarization, simplifying likelihood to single-polarization detection.

DQ

Data Quality — flags indicating whether detector data is suitable for science analysis. CAT0 = science mode, CAT1 = bad data (vetoed), CAT2 = marginal data (applied as veto windows).

event

A candidate gravitational-wave trigger produced by the likelihood pipeline. Includes time, frequency, sky position, SNR, and ranking statistic.

FAR

False Alarm Rate — expected rate of background triggers at or above a given ranking statistic. \(FAR(\rho^*) = N_{bkg}(\rho \ge \rho^*) / T_{bkg}\). Units: yr-1.

GWTC

Gravitational-Wave Transient Catalog — the LVK public catalog series of gravitational-wave detections. Public cWB pages provide waveform reconstruction and CED references for selected GWTC events.

HEALPix

Hierarchical Equal Area isoLatitude Pixelation — a scheme for pixelizing the sphere. healpix parameter controls sky resolution: \(N_{pix} = 12 \times 4^{healpix}\).

hrss

(\(h_{rss}\)) — root-sum-squared strain amplitude of a signal. Used as the amplitude measure for burst injections and efficiency.

hrss50 / hrss90

The \(h_{rss}\) at which 50% / 90% of injected signals are recovered. Standard figures of merit for search sensitivity.

injection

A simulated gravitational-wave signal added to detector data to measure search sensitivity. Defined by waveform parameters and sky location.

job

One unit of computation: a specific segment processed at a specific lag for a specific injection trial. Submitted to a cluster as an independent task.

job segment

A GPS time window containing detector data. A search is split into many segments that can run in parallel. Controlled by segLen, segMLS, segEdge.

lag

A time-shift applied to one detector’s data relative to others. Used to estimate the background: non-zero lags break real signal coincidences, producing only accidental triggers. Controlled by lagSize, lagStep, lagOff, lagMax.

lagOff

Offset for the first lag index. lagOff = 0 includes the zero-lag (physical coincidence). lagOff > 0 skips the first N lags.

LALInference

A legacy LIGO/Virgo Bayesian parameter-estimation toolkit used in older GWTC comparisons. Public cWB pages compare wavelet-based cWB reconstructions with LALInference modeled waveform reconstructions for some catalogs.

likelihood

The mathematical framework that evaluates how likely a cluster of pixels is to be a real GW signal rather than noise. Produces SNR, network correlation, and sky localization.

livetime

Total analyzed time used for background estimation. \(T_{bkg} = N_{jobs} \times (N_{lags} - 1) \times T_{seg}\).

MRA

Multi-Resolution Analysis — using wavelet transforms at multiple frequency resolutions to capture signals of different durations.

network

The set of gravitational-wave detectors used in a search. Common networks: LH (LIGO Hanford + Livingston), LHV (LIGO + Virgo).

pixel

A single time-frequency cell in the wavelet transform output. Excess power pixels are selected above a threshold and grouped into clusters.

postproduction

The pipeline that runs after all search jobs finish. Produces background estimates, XGBoost ranking, detection efficiency curves, and HTML reports.

progress

Per-job metadata (Parquet format) tracking which jobs/lags ran, how long each took, and whether they succeeded. Used by postproduction.

ranking statistic

The final score assigned to each event, combining multiple features (SNR, \(\chi^2\), correlation, etc.) via XGBoost. Used for FAR assignment and sensitivity.

rho

(\(\rho\)) — coherent network SNR. \(\rho = \sqrt{E_c - N_n}\) where \(E_c\) is coherent energy and \(N_n\) is null energy.

rho0

(\(\rho_0\)) — unsubtracted SNR: \(\rho_0 = \sqrt{E_c}\). Used when xgb_rho_mode: true.

segment

See job segment.

simulation

An injection study measuring how efficiently the search recovers signals. Requires matched injection truth tables.

sky mask

A restricted sky region for the likelihood scan. Used in targeted searches (e.g., GRB follow-up). Types: UniformAllSky, Patch, Fixed, Custom.

slag

Super lag — a segment-level time shift for multi-detector networks. Increases total jobs by factor of slagSize. Controlled by slagSize, slagMin, slagMax, slagOff.

SNR

Signal-to-Noise Ratio — measure of signal strength relative to background noise. See rho.

subnetwork cut

Per-sky-direction threshold cuts that remove pixel clusters unlikely to be astrophysical. Controlled by subnet, subcut, subrho.

supercluster

A merged group of nearby pixel clusters. The superclustering step determines which pixel groups are treated as a single candidate. Controlled by TFgap, Tgap, Fgap.

time-frequency map

The output of the wavelet transform: a 2D representation of signal power vs. time and frequency. Excess pixels are selected from this map.

trial / trial_idx

A group of injections sharing the same noise background within a job segment. trial_idx identifies which trial an injection belongs to.

WDM

Wilson-Daubechies-Meyer wavelet — the specific wavelet family used by cWB for time-frequency decomposition.

whitening

Normalizing the detector noise to have flat (white) frequency spectrum. Methods: wavelet, mesa, or mixed.

XGBoost

eXtreme Gradient Boosting — the machine-learning classifier used to combine multiple event features into a single ranking statistic.

XTalk

Cross-talk — coherent instrumental artifacts that mimic GW signals. pycWB includes XTalk identification and subtraction.

zero-lag

The unshifted (physical) coincidence between detectors where a real GW signal would appear. Excluded from background estimation.