pith. sign in

arxiv: 1907.04747 · v1 · pith:DYI4YBYOnew · submitted 2019-07-10 · 🌀 gr-qc · astro-ph.IM

Gravitational-wave parameter estimation with gaps in LISA: a Bayesian data augmentation method

Quentin Baghi , Ira Thorpe , Jacob Slutsky , John Baker , Tito Dal Canton , Natalia Korsakova , Nikos Karnesis This is my paper

Pith reviewed 2026-05-24 23:37 UTC · model grok-4.3

classification 🌀 gr-qc astro-ph.IM
keywords gravitational wave data analysisLISABayesian methodsdata gapsgalactic binariesparameter estimationdata augmentation
0
0 comments X

The pith

Bayesian data augmentation reintroduces missing LISA segments as auxiliary variables for consistent galactic binary parameter estimation.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces Bayesian data augmentation to address gaps in LISA interferometric data caused by various operational interruptions. The method treats the missing data segments as auxiliary variables within the Bayesian sampling of the posterior for source parameters. This provides a consistent statistical treatment that reduces noise leakage effects and enhances sampling efficiency. The approach is tested on the estimation of parameters for galactic binary sources under different gap scenarios, yielding results comparable to those from uninterrupted data.

Core claim

The paper establishes that by reintroducing the missing data as auxiliary variables in the sampling of the posterior distribution of astrophysical parameters, Bayesian data augmentation offers a statistically consistent method to handle gaps in LISA measurements. This mitigates problems such as noise leakage and increased computational complexity while improving sampling efficiency for the parameter estimation of galactic binaries.

What carries the argument

Bayesian data augmentation that reintroduces missing data segments as auxiliary variables in posterior sampling.

If this is right

  • Galactic binary parameters can be estimated accurately from gapped data without introducing bias from leakage.
  • The sampling process becomes more efficient than direct methods for incomplete data.
  • The method applies to various known gap patterns from LISA operations like laser switches or antenna re-pointing.
  • Posterior distributions remain reliable even when data is interrupted at different rates.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The technique may generalize to other LISA sources if their signals allow similar noise modeling.
  • Data analysis for future space-based observatories could benefit from built-in augmentation for handling random events.
  • Similar strategies might improve parameter estimation in other fields with intermittent observations, such as astronomy or signal processing.

Load-bearing premise

Gap locations and durations must be known in advance, and the underlying noise model must remain valid for the missing segments treated as auxiliary variables.

What would settle it

Simulate LISA data with known galactic binary signals and specific gaps, then check if the recovered posterior means and variances match those from the complete dataset within expected statistical fluctuations.

Figures

Figures reproduced from arXiv: 1907.04747 by Ira Thorpe, Jacob Slutsky, John Baker, Natalia Korsakova, Nikos Karnesis, Quentin Baghi, Tito Dal Canton.

Figure 1
Figure 1. Figure 1: FIG. 1: Segment of a simulated times series with [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2: Effective SNR as a function of the smoothing [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3: Result of PTMCMC sampling of the posterior distribution for ecliptic colatitude [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: In this figure we plot the periodograms of the com￾pleted data (black) and the gap-windowed data (gray), along with the estimated PSD using the windowing method (dotted brown), and using the DA method (dashed blue). In both cases the estimates are obtained by using the maximum a posteriori (MAP) estimator. The 3σ confidence intervals (light blue areas) are com￾puted using the sample variance of the posteri… view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Results of PSD estimations with gapped data, with five-day periodic gaps (left-hand side) and daily random [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 7
Figure 7. Figure 7: As in Fig. 6, they are ordered by decreasing sep [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6: Posterior distribution of frequencies [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7: Estimated Bayes factors [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
read the original abstract

By listening to gravity in the low frequency band, between 0.1 mHz and 1 Hz, the future space-based gravitational-wave observatory LISA will be able to detect tens of thousands of astrophysical sources from cosmic dawn to the present. The detection and characterization of all resolvable sources is a challenge in itself, but LISA data analysis will be further complicated by interruptions occurring in the interferometric measurements. These interruptions will be due to various causes occurring at various rates, such as laser frequency switches, high-gain antenna re-pointing, orbit corrections, or even unplanned random events. Extracting long-lasting gravitational-wave signals from gapped data raises problems such as noise leakage and increased computational complexity. We address these issues by using Bayesian data augmentation, a method that reintroduces the missing data as auxiliary variables in the sampling of the posterior distribution of astrophysical parameters. This provides a statistically consistent way to handle gaps while improving the sampling efficiency and mitigating leakage effects. We apply the method to the estimation of galactic binaries parameters with different gap patterns, and we compare the results to the case of complete data.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The manuscript introduces a Bayesian data augmentation method to handle gaps in LISA data for gravitational-wave parameter estimation. Missing data segments are introduced as auxiliary variables and jointly sampled with the astrophysical parameters (here, galactic binary parameters) under a stationary Gaussian noise model. The central claim is that this procedure is statistically equivalent to marginalization over the missing data, yielding a consistent posterior while improving sampling efficiency and reducing leakage compared to the complete-data case, with explicit comparisons across different gap patterns.

Significance. If the quantitative comparisons hold, the approach supplies a principled, internally consistent solution to a recurring practical problem in space-based GW data analysis. The explicit demonstration of equivalence to marginalization and the reported efficiency gains constitute a concrete contribution that could be adopted in LISA pipelines.

minor comments (2)
  1. Abstract: the claim of improved sampling efficiency and leakage mitigation is stated but not quantified; a single sentence summarizing the observed gains (e.g., effective sample size or wall-clock time) would strengthen the summary.
  2. Introduction or Methods: the assumption that gap locations and durations are known exactly should be stated explicitly as a modeling choice, together with a brief note on how the method would degrade if this assumption is violated.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of the manuscript, recognition of the statistical equivalence to marginalization, and recommendation for minor revision. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces Bayesian data augmentation to treat gaps in LISA data as auxiliary variables during posterior sampling for galactic binary parameters. This procedure is explicitly equivalent to marginalization over the missing segments under a stationary Gaussian noise model, which is a standard statistical identity and does not reduce any claimed benefit (efficiency gains or leakage mitigation) to a quantity defined by the method itself. No load-bearing step relies on self-citation of a uniqueness theorem, an ansatz smuggled from prior work, or a fitted parameter renamed as a prediction. The manuscript compares results directly to the complete-data case, providing an external benchmark. The derivation chain is therefore self-contained and independent of its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper rests on standard assumptions of Bayesian inference for time-series data and the validity of treating gaps as auxiliary variables; no free parameters, invented entities, or additional axioms are identifiable from the abstract.

axioms (1)
  • domain assumption Gap locations and durations are known and the noise statistics permit consistent augmentation without bias.
    Required for the method to mitigate leakage while remaining statistically consistent.

pith-pipeline@v0.9.0 · 5747 in / 1058 out tokens · 19398 ms · 2026-05-24T23:37:04.308992+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

47 extracted references · 47 canonical work pages · 11 internal anchors

  1. [1]

    We assume that this source emits a gravitational wave with strain polarizationsh+ and h× in the source frame

    Time domain model for gravitational-wave signal Let us consider a source located by radius r, colatitude θ and longitude φ in the solar-system barycentric ecliptic coordinate system. We assume that this source emits a gravitational wave with strain polarizationsh+ and h× in the source frame. We consider one of the interferometer arms of the LISA constella...

    work page

  2. [2]

    Noise model The noise model used to generate the data in this study is specified in Ref. [31], and can be written in relative frequency units as: SX(f) = 16 sin2 ( f f∗ )[ 2 ( 1 + cos2 ( f f∗ )) Saν(f) +Soν(f) ] , (28) where f∗≡ c 2πL ≈ 19 mHz is the LISA response trans- fer frequency and Saν(f) and Soν(f) are respectively the test-mass acceleration and th...

    work page

  3. [3]

    (6), it is preferable to have an analytic model of the response in the frequency domain

    Frequency domain model for gravitational-wave signal In order to efficiently compute the likelihood in Eq. (6), it is preferable to have an analytic model of the response in the frequency domain. This model can be approxi- mately derived in two steps. First, in the case of model (23) one can show that Eq. (21) can be re-written in the form of a linear combi...

    work page

  4. [4]

    five-day periodic gaps

    Parametrization of the noise power spectral density In a complex measurement like LISA, it is safer to es- timate the noise characteristics along with the signal, in order to avoid biases due to mismodeling of the noise PSD. Furthermore, we require some flexibility in the modeling, as the actual PSD may depart from the phys- ical model described by Eq. (28...

    work page

  5. [5]

    Laser Interferometer Space Antenna

    K. Danzmann, (2017), arXiv:1702.00786

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  6. [6]

    Aasi et al., Classical and Quantum Gravity 32, 074001 (2015)

    J. Aasi et al., Classical and Quantum Gravity 32, 074001 (2015)

    work page 2015

  7. [7]

    Acernese et al

    F. Acernese et al. , Classical and Quantum Gravity 32, 024001 (2015)

    work page 2015

  8. [8]

    Babak, J

    S. Babak, J. G. Baker, M. J. Benacquista, N. J. Cornish, and t. C. . Participants, Classical and Quantum Gravity 27, 084009 (2010). 18

    work page 2010

  9. [9]

    B. P. Abbott et al., Physical Review Letters 119, 161101 (2017), arXiv:1710.05832

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  10. [10]

    Armano et al

    M. Armano et al. , Physical Review Letters 120, 61101 (2018)

    work page 2018

  11. [11]

    S. E. Pollack, Classical and Quantum Gravity 21, 3419 (2004)

    work page 2004

  12. [12]

    The Effect of Data Gaps on LISA Galactic Binary Parameter Estimation

    J. Carr´ e and E. K. Porter, (2010), arXiv:arXiv:1010.1641v1

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  13. [13]

    Little and D

    R. Little and D. Rubin, Statistical analysis with missing data. (Second Edition.) (2002)

    work page 2002

  14. [14]

    M. J. Daniels and J. W. Hogan, Missing data in longitu- dinal studies : strategies for Bayesian modeling and sen- sitivity analysis (Chapman & Hall/CRC, 2008) p. 303

    work page 2008

  15. [15]

    J. R. Carpenter and M. G. Kenward, Multiple Imputation and its Application (John Wiley & Sons, Ltd, Chichester, UK, 2013)

    work page 2013

  16. [16]

    A. P. Dempster, N. M. Laird, and D. B. Rubin, (1976)

    work page 1976

  17. [17]

    M. A. Tanner and W. H. Wong, Journal of the American Statistical Association 82, 528 (1987)

    work page 1987

  18. [18]

    K. P. Murphy, Machine Learning: A Probabilistic Per- spective (1991) pp. 73–78,216–244, arXiv:0-387-31073-8

    work page 1991

  19. [19]

    Jaranowski and A

    P. Jaranowski and A. Kr´ olak, Living Reviews in Relativ- ity 15, 4 (2012)

    work page 2012

  20. [20]

    Luo and R

    Y. Luo and R. Duraiswami, Fast Near-GRID Gaussian Process Regression, Tech. Rep. (2013)

    work page 2013

  21. [21]

    J. R. Stroud, M. L. Stein, S. Lysen, S. is Ralph, and M. Otis Isham Professor, Bayesian and Maximum Like- lihood Estimation for Gaussian Processes on an Incom- plete Lattice, Tech. Rep. (2014) arXiv:arXiv:1402.4281v1

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  22. [22]

    Baghi, G

    Q. Baghi, G. M´ etris, J. Berg´ e, B. Christophe, P. Touboul, and M. Rodrigues, Phys. Rev. D 93, 122007 (2016)

    work page 2016

  23. [23]

    Hierarchical Nearest-Neighbor Gaussian Process Models for Large Geostatistical Datasets

    A. Datta, S. Banerjee, A. O. Finley, and A. E. Gelfand, Hierarchical Nearest-Neighbor Gaussian Process Mod- els for Large Geostatistical Datasets , Tech. Rep. (2016) arXiv:arXiv:1406.7343v2

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  24. [24]

    Whittle, Biometrika 41, 434 (1954)

    P. Whittle, Biometrika 41, 434 (1954)

    work page 1954

  25. [25]

    Priestley, Spectral analysis and time series , Probabil- ity and mathematical statistics No

    B. Priestley, Spectral analysis and time series , Probabil- ity and mathematical statistics No. vol. 1 to 2 (Academic Press, 1982)

    work page 1982

  26. [27]

    Fritz, I

    J. Fritz, I. Neuweiler, and W. Nowak, Mathematical Geo- sciences 41, 509 (2009)

    work page 2009

  27. [28]

    M. C. Edwards, R. Meyer, and N. Christensen, 10.1103/PhysRevD.92.064011

    work page doi:10.1103/physrevd.92.064011

  28. [29]

    Cutler, Physical Review D - Particles, Fields, Grav- itation and Cosmology 57, 7089 (1998), arXiv:9703068 [gr-qc]

    C. Cutler, Physical Review D - Particles, Fields, Grav- itation and Cosmology 57, 7089 (1998), arXiv:9703068 [gr-qc]

    work page 1998

  29. [30]

    N. J. Cornish and L. J. Rubbo, The LISA Response Func- tion, Tech. Rep. (2003)

    work page 2003

  30. [31]

    Blanchet, Living Reviews in Relativity 5, 3 (2002)

    L. Blanchet, Living Reviews in Relativity 5, 3 (2002)

    work page 2002

  31. [32]

    Tinto and S

    M. Tinto and S. V. Dhurandhar, Living Reviews in Rel- ativity 17, 6 (2014)

    work page 2014

  32. [33]

    J. W. Armstrong, F. B. Estabrook, and M. Tinto, The Astrophysical Journal 527, 814 (1999)

    work page 1999

  33. [34]

    Stebbins, LISA Science Requirements Document , Tech

    R. Stebbins, LISA Science Requirements Document , Tech. Rep

    work page

  34. [35]

    Krolak, M

    A. Krolak, M. Tinto, and M. Vallisneri, (2004), 10.1103/PhysRevD.70.022003 10.1103/Phys- RevD.76.069901, arXiv:0401108 [gr-qc]

    work page doi:10.1103/physrevd.70.022003 2004

  35. [36]

    N. J. Cornish and E. K. Porter, Class. Quantum Grav 22, 927 (2005)

    work page 2005

  36. [37]

    Mock LISA data challenge for the galactic white dwarf binaries

    A. B laut, S. Babak, and A. Kr´ olak, (2009), 10.1103/PhysRevD.81.063008, arXiv:0911.3020

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevd.81.063008 2009

  37. [40]

    T. B. Littenberg and N. J. Cornish, (2014), 10.1103/PhysRevD.91.084034, arXiv:1410.3852

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevd.91.084034 2014

  38. [41]

    Dynamic temperature selection for parallel-tempering in Markov chain Monte Carlo simulations

    W. Vousden, W. M. Farr, and I. Mandel, (2015), 10.1093/mnras/stv2422, arXiv:1501.05823

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stv2422 2015

  39. [42]

    emcee: The MCMC Hammer

    D. Foreman-Mackey, D. W. Hogg, D. Lang, and J. Good- man, (2012), 10.1086/670067, arXiv:1202.3665

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/670067 2012

  40. [43]

    T. B. Littenberg, Physical Review D 84, 063009 (2011)

    work page 2011

  41. [44]

    Markov Chain Monte Carlo Methods for Bayesian Data Analysis in Astronomy

    S. Sharma, Annual Review of Astronomy and Astro- physics 55, 1 (2017), arXiv:1706.01629v1

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  42. [45]

    Ogata, Numerische Mathematik 55, 137 (1989)

    Y. Ogata, Numerische Mathematik 55, 137 (1989)

    work page 1989

  43. [46]

    Gelman and X.-L

    A. Gelman and X.-L. Meng, Statistical Science , Tech. Rep. 2 (1998)

    work page 1998

  44. [47]

    Lartillot and H

    N. Lartillot and H. Philippe, Syst. Biol 55, 195 (2006)

    work page 2006

  45. [48]

    T. B. Littenberg and N. J. Cornish, 10.1103/Phys- RevD.80.063007

    work page doi:10.1103/phys-

  46. [49]

    Detecting Gravitational Wave Bursts with LISA in the presence of Instrumental Glitches

    T. Robson and N. J. Cornish, (2018), arXiv:1811.04490

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  47. [50]

    J. D. Romano and N. J. Cornish, Living reviews in rela- tivity 20, 2 (2017)

    work page 2017