Measuring the rate of glitches in interferometric gravitational wave detectors with a hierarchical Bayesian model 2026 Measuring the rate of glitches in interferometric gravitational wave detectors with a hierarchical Bayesian model – 21

Gregory Ashton, Colm Talbot, Andrew Lundgren, Ann-Kristin Malz, Joseph Areeda · 2026

Read Measuring the rate of glitches in interferometric gravitational wave detectors with a hierarchical Bayesian model 2026 Measuring the rate of glitches in interferometric gravitational wave detectors with a hierarchical Bayesian model – 21 by Gregory Ashton, Colm Talbot, Andrew Lundgren, Ann-Kristin Malz, Joseph Areeda (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-16039.txt.

Introduction

Ashton and four colleagues present a hierarchical Bayesian model to measure glitch rates in ground gravitational wave detectors. Glitches cut reach and can overlap signals, biasing cosmic readings. The model reaches into the low signal regime without Gaussian noise, if the population is modeled well. It covers short glitches only, via an antiglitch model. Applied to fourth run data, it gives time-tracked rates matching trigger counts without thresholds. Rates convert into a glitch chance, hinting withdrawn candidate GW230630_070659 is likely two glitches.

Catalog record

CREATORGregory Ashton, Colm Talbot, Andrew Lundgren, Ann-Kristin Malz, Joseph Areeda
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81645
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.16039

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Measuring the rate of glitches in interferometric gravitational wave detectors with a hierarchical Bayesian model 2026 Measuring the rate of glitches in interferometric gravitational wave detectors with a hierarchical Bayesian model – 21 Gregory Ashton, Colm Talbot, Andrew Lundgren, Ann-Kristin Malz, Joseph Areeda Abstract Ground-based gravitational wave detectors are now routinely surveying the dark Universe, finding hundreds of collisions between compact objects such as black holes and neutron stars. However, terrestrial non-Gaussian noise artefacts, commonly known as glitches, reduce the sensitivity to signals and can overlap signals, producing biased astrophysical inferences. We introduce a hierarchical Bayesian model to measure the glitch rate, which improves upon existing trigger-counting methods in its capacity to measure the rate down into the low signal-to-noise regime without contamination from the Gaussian noise background, provided the population is accurately modelled. The methodology builds on standard hierarchical inference, but includes several novel features, including hierarchical inference with quantile compression (HIQC), a generic approximation method for the hierarchical recycled likelihood, and a time-domain rate estimated by fitting basis functions. We validate the methodology using simulated data with injected glitches and then apply it to data from the fourth LIGO-Virgo-KAGRA observing run, demonstrating time-resolved inferences of the glitch rate over a 24 h 24\text{\,}\mathrm{h} period. The inferred glitch rate is consistent with estimates from trigger counts, but does not require an arbitrary threshold and provides a more fine-grained view of the temporal behaviour. Finally, we demonstrate how our individual-detector rate estimates can be transformed into a coincident glitch probability and utilise this to validate that the retracted gravitational-wave candidate GW230630_070659 is likely a pair of coincident glitches. 1 Introduction Since the inception of the LIGO ( 1 , LIGO:) , it has been known that non-Gaussian transient noise triggers are an unwanted feature of the interferometric strain and occur at a rate of a few times per hour ( 6 ) . These noise sources have now come to be known as glitches (see 19 , for a recent review) and they are also known ( 8 ; 9 ) to affect the Virgo ( 7 ) and KAGRA ( 10 ) detectors. In some cases, noise investigations reveal couplings between the glitches observed in the strain and environmental channels ( 35 ; 13 ; 33 ; 18 ) , enabling either mitigation of the source or vetoes to reduce the rate of false positives ( 29 ; 23 , see, e.g.) . While this has reduced the prevalence of some glitch classes, many have no environmental coupling, and as the Gaussian noise background is reduced, a constant background of glitches remains in the data from the LVK ( LVK ) detectors ( 39 ) . Glitches have two significant impacts. First, they reduce the sensitivity to astrophysical sources by increasing the rate of false positives and hence requiring a stricter threshold for detection. Second, they can overlap astrophysical sources, biasing the inferred properties ( 46 ; 25 ) . Therefore, significant effort is undertaken to investigate sources and understand the statistical properties of glitches ( 14 ) . In the most recent analysis of data from the first part of the fourth observing run (O4a), it was found that the rate of glitches was in the region of tens per hour ( 39 ) and similar rates were also reported during the third observing run ( 5 ) . Measuring the glitch rate is non-trivial. The standard approach is to analyse the strain data and auxiliary channels to find triggers using the Omicron software ( 36 ) . Applying a time-frequency decomposition, Omicron is applied in batches to all observed data, producing a list of triggers with corresponding measures of the SNR ( SNR ) and other properties. The rate can then be estimated as the number of events within the analysis duration, assuming Poissonian statistics. However, as with searching for any transient signal, in the absence of any glitches, Gaussian noise fluctuations will produce a distribution of triggers in SNR space; if the right-hand tail of this distribution overlaps the distribution of SNR produced by glitches, then the two cannot be unambiguously separated. Figure 1 : A histogram of the counts of Omicron trigger SNR over one day of data (2023-08-18) in the LIGO Livingston detector during O4a, along with time-frequency spectrograms ( 31 , created using GWpy :) of triggers selected from representative regions of the distribution: vertical dashed lines denote their SNR . The spectrograms cover a time span of 1\text{\,}\mathrm{s} and a frequency range from 20\text{\,}\mathrm{Hz} to 1\text{\,}\mathrm{kHz} . For the LVK detectors, the distribution of Omicron SNR for glitches does overlap the noise distribution. As an example, in Fig. 1 we show the distribution of Omicron trigger SNR from LLO ( LLO ) over a day in O4a down to a minimum SNR of 5 . From the time-frequency spectrograms of representative points in the distribution, we can observe that at the minimum SNR , the data is consistent with Gaussian noise. Meanwhile, triggers with larger SNR show distinct glitch behaviour. From the SNR distribution itself, there is no obvious threshold to disentangle the glitches from the Gaussian noise fluctuations. That the distributions overlap leads to a difficulty in measuring the rate of glitches: namely, a threshold must be selected to count glitches in a given time period. In 39 , SNR thresholds of 6.5 and 10 were used and these are common choices ( 5 , see, e.g. ) . Taking the day of O4a triggers from Fig. 1 , we extend the notion of measuring the rate at a fixed threshold, and in Fig. 2 , we plot the rate against the choice of SNR threshold. We see that above a threshold of 10 , the rate is reasonably stable, below 10^{-3} Hz. However, below an SNR of 10 , the rate markedly increases by orders of magnitude. This highlights the difficulty: choosing a conservative threshold will necessarily lead to an underestimate of the glitch rate, while choosing a more liberal threshold (e.g., an SNR of 6.5 ), the glitch rate may be overestimated due to false positives arising from Gaussian noise. Nevertheless, glitch rates based on an SNR threshold are the standard method used by the LVK to report glitch activity ( 39 ) . Moreover, the set of glitches obtained also forms the basis of further investigation. For example, 41 investigates the consistency of the measured glitch time with a Poisson distribution while 21 uses a clustering algorithm to understand the temporal behavior. Figure 2 : The estimated Poissonian glitch rate (and error) from Omicron triggers for one day of data during O4a as a function of the Omicron SNR threshold. Dashed lines mark the thresholds of 6.5 and 10 that are often used to determine the glitch rate. This work aims to introduce an alternative method for measuring the glitch rate using a hierarchical Bayesian framework. In part, this was motivated by work searching for the astrophysical background of merging compact binary coalescence ( 38 ) . In short, the method centres on the idea of applying full Bayesian parameter estimation to the full stream of data. Compared to the background search of 38 , our method differs

Read the complete text