Pith. sign in

REVIEW 2 major objections 6 minor 1 cited by

A Bayesian P-spline model of the inverse Cholesky factor recovers multivariate spectral matrices, and shows that realistic LISA noise needs the full cross-spectrum—not a diagonal AET assumption.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-11 12:47 UTC pith:SCSKWTEU

load-bearing objection Solid multivariate extension of Bayesian P-splines that cleanly shows AET-diagonal fails under realistic LISA MOSA imbalance; the RISE gap is a point-estimate result and holds up. the 2 major comments →

arxiv 2607.04833 v2 pith:SCSKWTEU submitted 2026-07-06 stat.ME astro-ph.IMgr-qc

Multivariate Bayesian P-spline estimation of spectral density matrices, with application to LISA TDI noise

Avi Vajpeyi , Renate Meyer , Patricio Maturana-Russel , Jianan Liu This is my paper
classification stat.ME astro-ph.IMgr-qc MSC 62M1562F1562G08
keywords Bayesian P-splinesspectral density matrixmultivariate time seriesWhittle likelihoodCholesky parametrisationLISA TDI noisesafe-Bayes temperingcross-spectra
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper builds a Bayesian way to estimate the full frequency-dependent noise covariance matrix of several channels at once, not just each channel’s power spectrum. It writes the inverse spectral matrix through a Cholesky factor at every frequency, then puts a smooth penalised B-spline prior on each real log-diagonal piece and each real and imaginary off-diagonal piece, so the matrix stays Hermitian positive definite by construction and the likelihood splits into independent channel regressions. On synthetic vector autoregressive series the method recovers both diagonals and cross-spectra, with credible intervals that cover near their nominal rate and error that falls as sample size grows. On simulated LISA time-delay interferometry data the same model agrees with a simple diagonal-AET fit when the six optical assemblies are identical, but when those assemblies have unequal noise the diagonal-AET assumption is off by more than an order of magnitude while the full multivariate fit still recovers the cross-spectral structure. That gap matters because LISA searches for stochastic backgrounds and parameter estimation both rely on a joint three-channel noise model.

Core claim

Parametrising the inverse spectral density matrix by its frequency-dependent Cholesky factor, placing independent P-spline priors on the log-diagonals and real/imaginary off-diagonals, and sampling a blocked, coarse-grained, η-tempered Whittle likelihood recovers both power and cross-spectra for multivariate stationary series. On realistic asymmetric LISA TDI simulations the unrestricted multivariate model reaches relative integrated squared (Frobenius) error around 10^{-3}, while a model forced to be diagonal in the AET basis stalls near 3.3×10^{-2}.

What carries the argument

Frequency-varying Cholesky factorisation of the inverse spectral matrix S(f)^{-1}=T* D^{-1} T, with independent penalised B-spline priors on each log-diagonal entry and each real and imaginary off-diagonal entry; this enforces Hermitian positive definiteness at every frequency and factorises the tempered Whittle likelihood into parallel per-channel regressions.

Load-bearing premise

Credible-interval calibration rests on a hand-chosen power η that down-weights the approximate Whittle likelihood; the paper has no automatic rule for setting that power from the data.

What would settle it

On a fresh asymmetric LISA realisation with a known analytic 3×3 spectrum, either the full multivariate posterior median still yields RISE ≳ 10^{-2} while a correctly specified transfer model stays near 10^{-3}, or the diagonal-AET model matches the full model to within ~10^{-3} RISE—either outcome would overturn the paper’s central LISA ordering claim.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • When LISA noise is equal-arm and equal-level, three independent univariate AET fits remain statistically adequate.
  • When per-MOSA noise levels differ, joint estimation of the full 3×3 TDI spectral matrix is required for unbiased stochastic-background and parameter-estimation pipelines.
  • With fixed tempering, posterior band widths on the diagonal PSDs contract roughly as the square root of observation time.
  • Blocking and coarse-graining can cut the number of frequency bins by large factors on smooth spectra without measurable loss of coverage or RISE.
  • The same Cholesky–P-spline construction applies to any stationary multivariate series in which cross-spectra matter.

Where Pith is reading between the lines

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

  • Published LISA stochastic-background forecasts that assume diagonal AET may be optimistic whenever real MOSA noise levels differ by the amounts already simulated here.
  • Replacing η-tempering with an analytically debiased Whittle likelihood (the route the paper itself flags) would remove the main hand-tuned calibration knob.
  • A piecewise-stationary extension of the same Cholesky–P-spline model could track constellation breathing and drifting link noise without abandoning the Whittle core.
  • If a richer variational family recovered calibrated bands, the existing SVI warm-start could become a production posterior at a fraction of full sampler cost.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper develops a Bayesian multivariate P-spline estimator for the frequency-dependent cross-spectral density matrix of stationary p-dimensional time series. The inverse spectrum is parametrised by a frequency-varying Cholesky factor so that Hermitian positive definiteness holds by construction; each real log-diagonal and real/imaginary off-diagonal Cholesky entry receives an independent penalised B-spline prior with adaptive knot placement. Inference uses a blocked, coarse-grained Whittle likelihood with safe-Bayes η-tempering, warm-started by low-rank SVI and sampled with NUTS. On a 3-D VAR(2) benchmark with closed-form spectra the method recovers diagonal and cross-spectral structure, achieves near-nominal 90% coverage across a (Nb, Nh) grid, and shows RISE decreasing with sample size. On public LISA TDI simulations, a nested H0 ⊂ H1 ⊂ H2 comparison shows that under symmetric MOSA noise an AET-diagonal model matches the full multivariate fit to ~10^{-3} in RISE, whereas under asymmetric per-MOSA noise the AET-diagonal model fails by more than an order of magnitude (~3.3×10^{-2} vs ~10^{-3}) while the unrestricted model recovers the cross-spectral structure.

Significance. If the results hold, the work supplies a practical, positive-definite nonparametric estimator for multivariate spectra that is directly usable for LISA TDI noise modelling, where cross-channel correlations matter for stochastic-background and parameter-estimation pipelines. Strengths that raise the contribution above a routine methods extension include: (i) the Cholesky factorisation that both enforces HPD and yields independent per-channel regressions that can be sampled in parallel; (ii) validation against external ground truth (closed-form VAR(2) spectra; analytic equal-link LDC and SEGWO/PyTDI references for LISA) rather than self-consistency; (iii) a clean nested H0/H1/H2 design that isolates the off-diagonal contribution; and (iv) open-source code with a pinned release and public LISA datasets, which makes the LISA RISE comparison reproducible. The bivariate Appendix B benchmark against VB/VNPC further situates the estimator among existing alternatives.

major comments (2)
  1. Sec. II C and Appendix C: safe-Bayes η is fixed at 0.5 for all LISA runs after a noise4a sweep that shows RISE flat to ≲5% while CI width contracts strongly with η. The abstract’s and Table II’s central RISE ordering on noise5a is therefore robust to this choice, but the reported ΔPSD values and the claim of “well-behaved” credible bands are η-dependent and are not accompanied by empirical coverage on LISA (only on the small VAR(2) grid at η=1). Please either (a) add a short noise5a η-sweep confirming that the H0/H1/H2 RISE gap is likewise flat, or (b) state explicitly that LISA ΔPSD should be read as a relative-precision diagnostic under a fixed tempering, not as calibrated frequentist coverage, and point practitioners to the open question of choosing η as a function of Nb Nh, K and spectral curvature.
  2. Sec. III B, Fig. 3 and the accompanying text: the unrestricted H2 model is said to “recover the cross-spectral structure,” yet the coherence panels show localised ripple near the TDI transfer-function nulls (~0.03, 0.06, 0.09 Hz) that the authors attribute to log-scale P-spline misfit at deep PSD dips. Because the headline claim is recovery of off-diagonal structure under asymmetric noise, please quantify how much of the residual RISE (~10^{-3}) is concentrated in those excised/near-null bands versus the rest of the band, so that readers can judge whether the residual is scientifically negligible for SGWB/PE use cases.
minor comments (6)
  1. Eq. (4) and the surrounding paragraph: the coherence formula is written with a line-break artefact (“/github(4)” style) in the source; ensure the published equation is clean and that |Cij|∈[0,1] is stated once without duplication.
  2. Table I: wall-clock times are essentially constant across a 32-fold change in Nc; a one-sentence note that at this (K,p,n) the cost is dominated by NUTS warm-up/JIT/SVI rather than the O(p² Nc K) likelihood would help readers extrapolate to larger problems (as you do later for LISA).
  3. Sec. II D: the adaptive knot rule (denoised periodogram gradient → quantile placement) is important for LISA’s steep low-frequency rise; a short pseudocode box or pointer to the exact function in the pinned v0.1.0 release would aid reimplementation.
  4. Appendix B, Table IV: the L2 metric is un-normalised integrated Frobenius error, whereas the main text uses RISE; state the relation once so that the bivariate numbers can be compared in scale to Table I.
  5. Notation Table III is very useful; consider moving a one-line pointer to it into the first paragraph of Sec. II rather than only in the appendix lead-in.
  6. Typos / polish: “aetnoise” spacing in the abstract; “MUL TIV ARIA TE” and “APPLICA TION” header spacing; “Gr¨ unwald” encoding; and a few “/github” artefacts left in the manuscript body should be cleaned before production.

Circularity Check

0 steps flagged

No significant circularity: central RISE claims are scored against external closed-form VAR(2) and analytic LISA (LDC/SEGWO) references under nested models, not against self-fitted targets.

full rationale

The paper’s derivation chain is a standard Bayesian nonparametric construction (Cholesky of S^{-1}, independent P-spline priors on log-diagonals and Re/Im off-diagonals, blocked/coarse-grained Whittle likelihood, optional η-tempering, NUTS from SVI init). Positive-definiteness is enforced by construction of the Cholesky factorisation (Rosen & Stoffer; Hu & Prado), not by circular appeal to the data. Validation is against external ground truth: the theoretical one-sided VAR(2) spectrum S(f)=2/fs H(f)ΣH(f)* with known A1,A2,Σ; and for LISA, closed-form equal-link LDC transfer (noise4a) and SEGWO/PyTDI projection of measured per-MOSA ASDs (noise5a). The headline order-of-magnitude RISE gap on noise5a (H0 ~3.3e-2 vs H2 ~1e-3) is a nested-model comparison H0⊂H1⊂H2 against those analytic references. Self-citations to the authors’ univariate P-spline work and the bivariate Liu et al. benchmark supply prior methodology and a secondary accuracy check; they are not used as the validation target of the central claim. Appendix C shows matrix RISE is essentially flat in η (≲5% variation), so the hand-chosen η=0.5 affects CI width, not the point-estimate RISE ordering. No step reduces a claimed prediction to a fitted input by definition, and no uniqueness theorem is imported from the authors to forbid alternatives. Score 0 is therefore appropriate.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The method rests on standard asymptotic spectral theory plus a small set of modelling choices (stationarity, Whittle, independent P-spline components, hand-chosen η). No new physical entities are postulated; free parameters are the usual spline and tempering knobs that control smoothness and calibration.

free parameters (4)
  • safe-Bayes tempering η = 0.5 (LISA); 1.0 (VAR(2))
    Fixed at 0.5 for all LISA runs after a discrete sensitivity sweep; the paper notes no principled selection rule yet exists.
  • number of B-spline basis functions K (and Kθ) = K=10 (sim); K=100, Kθ=2 (LISA H1)
    Chosen by hand (K=10 for VAR(2), K=100 / Kθ=2 for LISA) to balance flexibility and smoothness.
  • blocking Nb and coarse-grain Nh = Nb≈4–52; Nc=1024
    Nb chosen so blocks are ~1 week for LISA; Nh set to keep Nc=1024; both affect effective sample size and likelihood curvature.
  • hierarchical prior hyperparameters αϕ, βϕ, αν, βν = all = 1
    Defaulted to 1 without data-driven tuning; control the strength of the roughness penalty.
axioms (4)
  • domain assumption Wide-sense stationarity of the observed multivariate series so that the spectral density matrix S(f) is well-defined and the DFT coefficients are asymptotically independent complex Gaussian.
    Stated in Sec. II A; both LISA datasets are stationary by construction, but real LISA is acknowledged to be non-stationary.
  • domain assumption The blocked, coarse-grained Whittle likelihood (complex Wishart) is an adequate approximation after ENBW and η corrections.
    Sec. II A–C; the paper itself notes mild misspecification and introduces η-tempering precisely because of it.
  • ad hoc to paper Independent hierarchical Gaussian smoothing priors on each real/imaginary Cholesky component, with integral second-derivative penalty matrices.
    Sec. II D–E; the independence across components is a modelling choice that yields posterior factorisation and parallel sampling.
  • standard math Hermitian positive-definiteness of S(f) is enforced by parametrising S(f)^{-1} via its Cholesky factor with positive diagonal entries.
    Sec. II B; standard linear-algebra fact used by Rosen & Stoffer and Hu & Prado.

pith-pipeline@v1.1.0-grok45 · 26892 in / 3023 out tokens · 27023 ms · 2026-07-11T12:47:56.953162+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Multivariate Bayesian P-spline estimation of spectral density matrices, with application to LISA TDI noise." pith.science (2026). https://pith.science/paper/SCSKWTEU

@misc{pith2026260704833,
  author       = {Pith},
  title        = {Pith review of: Multivariate Bayesian P-spline estimation of spectral density matrices, with application to LISA TDI noise},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SCSKWTEU}},
  note         = {Machine review of arXiv:2607.04833}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We present a Bayesian P-spline method for estimating the frequency-dependent cross-spectral density matrix of stationary multivariate time series. The inverse spectral matrix is parametrised through its frequency-varying Cholesky decomposition, which guarantees Hermitian positive definiteness at every frequency. Each real log-diagonal entry and each real and imaginary off-diagonal entry is given an independent penalised B-spline prior that controls smoothness. Inference uses a blocked, coarse-grained Whittle likelihood with safe-Bayes $\eta$-tempering to stabilise posterior calibration, sampled by the No-U-Turn Sampler from a variational initialisation. On synthetic VAR(2) benchmarks with known ground truth, the method recovers both diagonal and cross-spectral structure, attains near-nominal credible-interval coverage, and achieves a relative integrated squared (Frobenius) error (RISE) that decreases with sample size. We then apply the method to publicly released simulated LISA time-delay interferometry (TDI) data in two noise configurations. In the idealised symmetric case, the full multivariate model and a reduced model that assumes a diagonal AET noise covariance agree to within $\sim10^{-3}$ in RISE. Under realistic noise that is asymmetric across the six Movable Optical Sub-Assemblies (MOSAs), the AET-diagonal assumption fails by more than an order of magnitude in RISE ($\sim\!3.3\!\times\!10^{-2}$ versus $\sim\!10^{-3}$), whereas the full multivariate model recovers the cross-spectral structure.

Figures

Figures reproduced from arXiv: 2607.04833 by Avi Vajpeyi, Jianan Liu, Patricio Maturana-Russel, Renate Meyer.

Figure 1
Figure 1. Figure 1: FIG. 1. Construction of the blocked, coarse-grained periodogram. The [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Representative 3 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Posterior summaries of the LISA TDI [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Safe-Bayes [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Bayesian nonparametric estimation of correlated gravitational wave detector network noise using matrix-gamma process priors

    gr-qc 2026-07 conditional novelty 6.0

    A Bayesian nonparametric model estimates the full cross-channel noise spectrum of LISA/ET detectors with positive-definite spectral matrices and MCMC-computed uncertainties.

Reference graph

Works this paper leans on

37 extracted references · 3 canonical work pages · cited by 1 Pith paper

  1. [1]

    Deep analysis group

    model the diagonal PSDs and the real and imaginary parts of the cross-spectra ofS(f) as smooth fractional de- viations from a design spectrum, each represented by a natural cubic spline. They note that this construction is not guaranteed to yield a positive-definite matrix away from the design point, and that it is well suited to their local Fisher-matrix...

    2000

  2. [2]

    T. B. Littenberg and N. J. Cornish, Bayesian inference for spectral estimation of gravitational wave detector noise, Physical Review D91, 084034 (2015)

    2015

  3. [3]

    P. A.-S. et al., Laser interferometer space antenna (2017), arXiv:1702.00786

    Pith/arXiv arXiv 2017

  4. [4]

    Kirch, M

    C. Kirch, M. C. Edwards, A. Meier, and R. Meyer, Be- yond Whittle: Nonparametric Correction of a Parametric Likelihood with a Focus on Bayesian Time Series Analy- sis, Bayesian Analysis14, 1037 (2019)

    2019

  5. [5]

    J. Liu, A. Vajpeyi, R. Meyer, K. Janssens, J. E. Lee, et al., Variational inference for correlated gravitational wave detector network noise, Phys. Rev. D111, 062003 14 Symbol Meaning Time series & sampling T,∆ t, f s Total observation duration, sampling interval, sampling frequency;T=n∆ t, ∆t = 1/fs. fN y Nyquist frequency,f N y =f s/2. n, pNumber of tim...

    2025

  6. [6]

    M. C. Edwards, R. Meyer, and N. Christensen, Bayesian nonparametric spectral density estimation using B-spline priors, Statistics and Computing29, 67 (2019)

    2019

  7. [7]

    P. H. C. Eilers and B. D. Marx, Flexible smoothing with B-splines and penalties, Statistical Science11, 89 (1996)

    1996

  8. [8]

    Maturana-Russel and R

    P. Maturana-Russel and R. Meyer, Bayesian spec- tral density estimation using P-splines with quantile- based knot placement, Computational Statistics36, 2055 (2021)

    2055

  9. [9]

    Aimen, P

    N. Aimen, P. Maturana-Russel, A. Vajpeyi, N. Chris- tensen, and R. Meyer, Bayesian power spectral density es- timation for LISA noise based on penalized splines with a parametric boost, Physical Review D113, 024022 (2026)

    2026

  10. [10]

    Baghi, N

    Q. Baghi, N. Karnesis, J.-B. Bayle, M. Besan¸ con, 15 and H. Inchausp´ e, Uncovering gravitational-wave back- grounds from noises of unknown shape with LISA, Jour- nal of Cosmology and Astroparticle Physics2023(04), 066

  11. [11]

    Santini, M

    A. Santini, M. Muratore, J. Gair, and O. Hartwig, Flex- ible, GPU-accelerated approach for the joint character- ization of LISA instrumental noise and stochastic grav- itational wave backgrounds, Phys. Rev. D112, 084050 (2025), arXiv:2507.06300 [gr-qc]

    arXiv 2025

  12. [12]

    Muratore,Instrumental modelling and noise reduction algorithms for the Laser Interferometer Space Antenna, Ph.D

    M. Muratore,Instrumental modelling and noise reduction algorithms for the Laser Interferometer Space Antenna, Ph.D. thesis, Gottfried Wilhelm Leibniz Universit¨ at Han- nover (2021)

    2021

  13. [13]

    Cireddu, M

    F. Cireddu, M. Wils, I. C. F. Wong, P. T. H. Pang, T. G. F. Li,et al., Likelihood for a network of gravitational-wave detectors with correlated noise, Phys. Rev. D110, 104060 (2024)

    2024

  14. [14]

    Hartwig, M

    O. Hartwig, M. Lilley, M. Muratore, and M. Pieroni, Stochastic gravitational wave background reconstruction for a nonequilateral and unequal-noise lisa constellation, Phys. Rev. D107, 123531 (2023)

    2023

  15. [15]

    Muratore, J

    M. Muratore, J. Gair, and L. Speri, Impact of the noise knowledge uncertainty for the science exploitation of cos- mological and astrophysical stochastic gravitational wave background with lisa, Phys. Rev. D109, 042001 (2024)

    2024

  16. [16]

    Rosen and D

    O. Rosen and D. S. Stoffer, Automatic estimation of mul- tivariate spectra via smoothing splines, Biometrika94, 335 (2007)

    2007

  17. [17]

    Hu and R

    Z. Hu and R. Prado, Fast Bayesian inference on spectral analysis of multivariate stationary time series, Computa- tional Statistics & Data Analysis178, 107596 (2023)

    2023

  18. [18]

    Gr¨ unwald and T

    P. Gr¨ unwald and T. van Ommen, Inconsistency of Bayesian inference for misspecified linear models, and a proposal for repairing it, Bayesian Analysis12, 1069 (2017)

    2017

  19. [19]

    M. D. Hoffman, D. M. Blei, C. Wang, and J. Pais- ley, Stochastic variational inference, Journal of machine learning research (2013)

    2013

  20. [20]

    M. D. Hoffman, A. Gelman,et al., The no-u-turn sam- pler: adaptively setting path lengths in hamiltonian monte carlo., J. Mach. Learn. Res.15, 1593 (2014)

    2014

  21. [21]

    Y. Liu, C. Kirch, J. E. Lee, and R. Meyer, A nonparamet- rically corrected likelihood for bayesian spectral analysis of multivariate time series (2024)

    2024

  22. [22]

    J. P. Grainger, A. M. Sykulski, K. Ewans, H. F. Hansen, and P. Jonathan, A multivariate pseudo-likelihood ap- proach to estimating directional ocean wave models, Journal of the Royal Statistical Society Series C: Applied Statistics72, 544 (2023)

    2023

  23. [23]

    M. P. Wand and J. T. Ormerod, On semiparametric re- gression with o’sullivan penalized splines, Australian & New Zealand Journal of Statistics50, 179 (2008)

    2008

  24. [24]

    Bradbury, R

    J. Bradbury, R. Frostig, P. Hawkins, M. J. Johnson, C. Leary,et al., JAX: composable transformations of Python+NumPy programs (2018)

    2018

  25. [25]

    D. Phan, N. Pradhan, and M. Jankowiak, Composable effects for flexible and accelerated probabilistic program- ming in pyro, arXiv preprint arXiv:1912.11554 (2019)

    Pith/arXiv arXiv 1912

  26. [26]

    D. P. Kingma and J. Ba, Adam: A method for stochastic optimization, in3rd International Conference on Learn- ing Representations (ICLR)(2015) arXiv:1412.6980 [cs.LG]

    Pith/arXiv arXiv 2015

  27. [27]

    Gelman and D

    A. Gelman and D. B. Rubin, Inference from Iterative Simulation Using Multiple Sequences, Statistical Science 7, 457 (1992)

    1992

  28. [28]

    Kumar, C

    R. Kumar, C. Carroll, A. Hartikainen, and O. Mar- tin, Arviz a unified library for exploratory analysis of bayesian models in python, Journal of Open Source Soft- ware4, 1143 (2019)

    2019

  29. [29]

    Bayle, LISA SGWB Dataset (noise-4a) (2025), https://zenodo.org/doi/10.5281/zenodo.15698080

    J.-B. Bayle, LISA SGWB Dataset (noise-4a) (2025), https://zenodo.org/doi/10.5281/zenodo.15698080

    work page doi:10.5281/zenodo.15698080 2025

  30. [30]

    Bayle, Lisa sgwb dataset (noise-5a), To be assigned (2026)

    J.-B. Bayle, Lisa sgwb dataset (noise-5a), To be assigned (2026)

    2026

  31. [31]

    Bayle, O

    J.-B. Bayle, O. Hartwig, and M. Staab, Lisa instrument (2024)

    2024

  32. [32]

    Le Jeune, S

    M. Le Jeune, S. Babak, Q. Baghi, J.-B. Bayle, E. Castelli, and N. Korsakova, Lisa data challenge spritz (ldc2b), 10.5281/zenodo.7436568 (2022)

    work page doi:10.5281/zenodo.7436568 2022

  33. [33]

    Staab, J.-B

    M. Staab, J.-B. Bayle, and O. Hartwig, PyTDI (2025), https://zenodo.org/doi/10.5281/zenodo.6351736

    work page doi:10.5281/zenodo.6351736 2025

  34. [34]

    T. A. Prince, M. Tinto, S. L. Larson, and J. W. Armstrong, LISA optimal sensitivity, Phys. Rev. D66, 122002 (2002), arXiv:gr-qc/0209039 [gr-qc]

    Pith/arXiv arXiv 2002

  35. [35]

    Bayle and O

    J.-B. Bayle and O. Hartwig, SEGWO: Sensitivity esti- mator for gravitational-wave observatories

  36. [36]

    M. L. Katzet al., The LISA global fit, arXiv preprint arXiv:2404.12571 (2024)

    Pith/arXiv arXiv 2024

  37. [37]

    A. M. Sykulski, S. C. Olhede, A. P. Guillaumin, J. M. Lilly, and J. J. Early, The debiased Whittle likelihood, Biometrika106, 251 (2019)

    2019