Pith. sign in

REVIEW 3 major objections 1 minor 46 references

Optimal Interference Signal for Masking an Acoustic Source

T0 review · 3 major / 1 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read For spherically symmetric waves, analytical interference signals can cancel an acoustic source in a target region, hiding it from sensors.

desk verdict The manuscript body is a different paper, so the acoustic-masking claims are unauditable; desk reject this version and ask for the real text. read the letter →

arxiv 2508.15023 v1 pith:G2CAT4NE submitted 2025-08-20 math.AP cs.SDeess.AS

Hongyun Wang , Hong Zhou This is my paper
classification math.APcs.SDeess.AS MSC 35L0535B10
keywords acousticmaskinginterferencesignalsforcedwaveequationsphericalsymmetryquasi-steadyperiodicsolutionsself-maskingpoint-forcesourcesunderseastealth
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper aims to turn acoustic masking into a design problem that can be solved on paper. For the three-dimensional forced wave equation with spherical symmetry, it derives analytical quasi-steady periodic solutions for interference signals placed near a source, then shows how to superpose them so the combined field has near-zero amplitude in a target region where detection sensors sit. It also identifies a self-masking effect: a source with the right spatial forcing profile masks itself outside its own footprint without needing any added signal. The same construction is used to analyze one or two point-force interferers, and a numerical method extends the approach to sources without spherical symmetry. If the idealizations hold, this provides a first-principles recipe for undersea communication security and vehicle stealth.

What carries the argument

The load-bearing object is the family of quasi-steady periodic spherically symmetric solutions to the three-dimensional forced wave equation. Linearity lets these solutions be superposed: the interference field is chosen so that, in the target region, it is as close as possible to the negative of the source field. The point-force solutions then act as the controllable building blocks for the cancellation recipe.

What would settle it

In a water tank with controlled boundaries, drive a small acoustic source and the optimal one- or two-point-force interference computed by the paper, then measure the RMS pressure in the target region; if the measured residual is not substantially lower than nearby non-optimal placements, the optimal-signal rule is wrong.

Watch

Extended reading notes

Core claim

The paper's central claim is that masking can be solved at the level of the acoustic field, not by trial-and-error noise. Working with the three-dimensional forced wave equation under spherical symmetry, it derives quasi-steady periodic solutions describing the field of an interference signal. Since the equation is linear, the total field is a sum: source field plus interference field. The paper chooses the interference so that the sum has minimum amplitude in a given target region, and turns this into explicit recipes for one or two point forces placed near the source. A self-masking effect also emerges: a source with a spatial forcing profile confined to a footprint masks itself outside th

Load-bearing premise

The analytical results assume a homogeneous, lossless, linear acoustic medium and steady periodic signals with spherical symmetry; real undersea environments have boundaries, refraction, attenuation, and transient sources.

Editorial extensions

If this is right

  • Acoustic masking can be designed from first principles: given a source and a target region, the interference amplitudes and phases are obtained analytically rather than by heuristics.
  • One or two point forces near the source can serve as practical masking devices, and their optimal settings are computable from the paper's superposition rule.
  • For sources with the right spatial forcing profile, self-masking means no external interference is needed; the source is already invisible outside its footprint.
  • The numerical method extends the same design idea to sources that are not spherically symmetric, covering more realistic vehicle or communication sources.
  • The results give a concrete path toward undersea communication security and stealth, since the masking signals are explicit fields rather than generic jamming noise.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension is to treat multiple target regions or moving sensors by optimizing a sum of residual amplitudes over the same superposition, a problem not addressed explicitly in the paper.
  • Because the analytical results assume a homogeneous lossless medium, a practical version of the rule would minimize the worst-case residual under uncertain sound speed or boundaries; the paper leaves that question implicit.
  • The self-masking condition, if read backward, becomes a design constraint on the source's forcing profile: shaping a source so it is self-masking might be easier than deploying interference.
  • The numerical method could be tested against the analytic spherically symmetric solutions as a benchmark, providing a direct check that the solver implements the same design principle.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 1 minor

Summary. The manuscript as supplied carries the title 'Optimal Interference Signal for Masking an Acoustic Source' and an abstract promising analytical quasi-steady periodic solutions for the three-dimensional forced wave equation with spherical symmetry, a superposition-based masking framework, a self-masking phenomenon, optimization of one- and two-point-force interference signals, and an efficient numerical method for the non-symmetric case. The full text, however, is not this paper. It is Arman Oganisian's Bayesian causal inference manuscript 'Untangling Sample and Population Level Estimands in Bayesian Causal Computation' (arXiv:2508.15016v3), whose Sections 1–9 and Appendices A–E develop ITE/SATE/CATE/PATE estimands, identification, MCMC sampling, and the Bayesian g-formula. None of the acoustic-masking content advertised in the abstract appears anywhere in the supplied body. The central claims of the masking paper are therefore unsupported by any derivable evidence in this submission and cannot be audited for correctness, internal consistency, or numerical validation.

Significance. If the claimed results existed, they could constitute a useful first-principles framework for designing interference signals that minimize acoustic residual amplitude in a target region, with potential applications to undersea acoustic privacy and surveillance countermeasures. However, the significance cannot be assessed from the present deliverable: the paper provides no forced-wave-equation derivation, no spherical symmetry reduction, no quasi-steady periodic ansatz, no objective-function minimization, no point-force optimization results, and no numerical experiments. There are also no machine-checked proofs, reproducible code, or parameter-free derivations to credit. On the evidence supplied, the submission is not the paper it claims to be, so no scientific claim is currently established.

major comments (3)
  1. [Full text, all sections] The supplied body is an unrelated manuscript: A. Oganisian, 'Untangling Sample and Population Level Estimands in Bayesian Causal Computation' (arXiv:2508.15016v3). It contains Eqs. (1)–(11) defining causal estimands and posterior distributions, MCMC algorithms, and Stan code. The acoustic-masking abstract's promised content—the 3D forced wave equation, spherical symmetry, quasi-steady periodic solutions, superposition, residual-amplitude minimization, self-masking, and one- or two-point-force optimization—is entirely absent. This is a load-bearing failure: the central claim can neither be verified nor falsified from the manuscript.
  2. [Equations (1)–(11)] The only equations in the deliverable belong to Bayesian causal inference (e.g., the joint posterior in Eq. (2), the factorization in Eq. (3), the g-formula in Eq. (8), and Monte Carlo approximations in Eqs. (9)–(10)). None of them relates to wave propagation, masking, or acoustic source suppression. There is no derivation of quasi-steady periodic solutions, no error or convergence analysis, and no numerical method for the 3D wave equation. The assertions in the abstract therefore rest on no auditable mathematical content.
  3. [Abstract] Even considered in isolation, the abstract states that 'we develop an efficient numerical method for solving the 3D wave equation' for the general nonsymmetric case, but the submission contains no numerical solver, no validation, no complexity discussion, and no accuracy assessment. Since the promised theoretical core is also missing, the manuscript cannot be meaningfully revised in place; it would need to be replaced by a different document.
minor comments (1)
  1. [Bibliographic metadata] The author affiliation, keywords (62C10, 62F15, 62G08), and reference list are those of the Bayesian statistics paper and have no connection to the masking title. If the correct manuscript exists, the metadata and body must be reconciled; as presented, the submission is internally inconsistent at the level of its most basic identification.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity can be established: the supplied full text is a different paper, so the claimed derivation chain is absent rather than circular.

full rationale

The abstract of arXiv:2508.15023 promises analytical quasi-steady periodic masking solutions for the 3D forced wave equation, superposition principles, point-force optimization, a self-masking phenomenon, and a numerical method for the general case. The supplied full text, however, is Arman Oganisian's Bayesian causal inference manuscript (arXiv:2508.15016v3), containing none of these derivations: there is no forced wave equation, no spherically symmetric ansatz, no superposition formula, no residual-amplitude objective, and no numerical scheme for the 3D wave equation. Under the reviewing rule, this mismatch is in-scope evidence, but it is not circularity. Circularity requires a specific load-bearing step that reduces, by the paper's own equations or by self-citation, to its own inputs. No such step can be quoted because the claimed derivation chain is absent from the provided text. The structural caution that 'any optimal interference signal is, by construction, the minimizer of the residual-amplitude objective' is a general observation about optimization formulations, not a demonstrated reduction in this paper. Likewise, the absence of validation against external measurements or independent predictions is a support/audit concern, not a circularity defect. Since no quotation exhibits Eq. X = Eq. Y by construction, no fitted parameter renamed as a prediction, and no load-bearing self-citation chain, the honest finding is no circularity (score 0). The paper's central claim may be unsupported in the supplied text, but unsupported and circular are different; only the former can be asserted here.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The ledger reflects the abstract only, because the supplied full text is an unrelated manuscript. No new physical entities are introduced: interference signals are standard acoustic sources and 'self-masking' is a described phenomenon rather than a new entity. The free parameters are the point-force degrees of freedom that an optimization procedure selects. The axioms are the idealizations of a homogeneous linear medium, steady periodic signals, and spherical symmetry that underpin the claimed analytical results.

free parameters (1)
  • Point-force interference parameters (amplitude, phase, position) for the one and two point-force cases = not stated in abstract
    The abstract states the paper 'analyze[s] and optimize[s] the performance of using one or two point-forces deployed near the acoustic source'. These are the optimization variables of the framework, so any reported masking performance depends on their chosen values.
assumptions (3)
  • domain assumption The 3D forced wave equation in a homogeneous, linear medium is the correct model for acoustic propagation in the masking scenario.
    The abstract builds everything on 'the three-dimensional (3D) forced wave equation with spherical symmetry', ignoring boundaries, stratification, and attenuation that real undersea environments exhibit.
  • domain assumption Quasi-steady periodic solutions capture the relevant masking behavior.
    The abstract says the authors 'derive analytical quasi-steady periodic solutions', which restricts the analysis to steady periodic signals and may not cover transient, broadband, or time-varying acoustic sources.
  • domain assumption Spherical symmetry is a valid canonical case for the analytical solutions, with the general non-symmetric case left to the numerical method.
    The abstract limits the analytical derivation to the spherically symmetric forcing case and states that 'the general case where the spatial forcing profile of the acoustic source lacks spherical symmetry' is handled by 'an efficient numerical method'. The analytical guarantees therefore do not extend directly to general source shapes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimal Interference Signal for Masking an Acoustic Source." pith.science (2026). https://pith.science/paper/G2CAT4NE

@misc{pith2026250815023,
  author       = {Pith},
  title        = {Pith review of: Optimal Interference Signal for Masking an Acoustic Source},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G2CAT4NE}},
  note         = {Machine review of arXiv:2508.15023}
}
read the original abstract

In an environment where acoustic privacy or deliberate signal obfuscation is desired, it is necessary to mask the acoustic signature generated in essential operations. We consider the problem of masking the effect of an acoustic source in a target region where possible detection sensors are located. Masking is achieved by placing interference signals near the acoustic source. We introduce a theoretical and computational framework for designing such interference signals with the goal of minimizing the residual amplitude in the target region. For the three-dimensional (3D) forced wave equation with spherical symmetry, we derive analytical quasi-steady periodic solutions for several canonical cases. We examine the phenomenon of self-masking where an acoustic source with certain spatial forcing profile masks itself from detection outside its forcing footprint. We then use superposition of spherically symmetric solutions to investigate masking in a given target region. We analyze and optimize the performance of using one or two point-forces deployed near the acoustic source for masking in the target region. For the general case where the spatial forcing profile of the acoustic source lacks spherical symmetry, we develop an efficient numerical method for solving the 3D wave equation. Potential applications of this work include undersea acoustic communication security, undersea vehicles stealth, and protection against acoustic surveillance.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 33 canonical work pages

  1. [1]

    Causal inference in high dimensions: A marriage between bayesian modeling and good frequentist properties.����������, 78:100–114, 3 2022

    Joseph Antonelli, Georgia Papadogeorgou, and Francesca Dominici. Causal inference in high dimensions: A marriage between bayesian modeling and good frequentist properties.����������, 78:100–114, 3 2022. 10.1111/BIOM.13417

    work page doi:10.1111/biom.13417 2022

  2. [2]

    Conservative inference for counterfactuals.������� �� ������ ���������, 13(1):20230071, 2025

    Sivaraman Balakrishnan, Edward Kennedy, and Larry Wasserman. Conservative inference for counterfactuals.������� �� ������ ���������, 13(1):20230071, 2025. 10.1515/jci-2023-0071

    work page doi:10.1515/jci-2023-0071 2025

  3. [3]

    Targeted estimation and inference for the sample average treatment effect in trials with and without pair-matching.���������� �� ��������, 35(21):3717– 3732, 2016

    Laura B Balzer, Maya L Petersen, Mark J van der Laan, and Search Collaboration. Targeted estimation and inference for the sample average treatment effect in trials with and without pair-matching.���������� �� ��������, 35(21):3717– 3732, 2016. 10.1002/sim.6965

    work page doi:10.1002/sim.6965 2016

  4. [4]

    Chipman, Edward I

    Hugh A. Chipman, Edward I. George, and Robert E. McCulloch. BART: Bayesian additive regression trees.��� ��� ���� �� ������� ����������, 4(1):266 – 298, 2010. 10.1214/09-AOAS285. URL https://doi.org/10.1214/09-AOAS285

    doi:10.1214/09-aoas285 2010

  5. [5]

    Chapman and Hall/CRC, 2023

    Michael J Daniels, Antonio Linero, and Jason Roy.�������� �������������� ��� ������ ��������� ��� ������� ����. Chapman and Hall/CRC, 2023

    work page 2023

  6. [6]

    Causal Inference: A Missing Data Perspective.����������� �������, 33(2):214 – 237, 2018

    Peng Ding and Fan Li. Causal Inference: A Missing Data Perspective.����������� �������, 33(2):214 – 237, 2018. 10.1214/18-STS645

    doi:10.1214/18-sts645 2018

  7. [7]

    Miratrix

    Peng Ding, Xinran Li, and Luke W. Miratrix. Bridging finite and super population causal inference.������� �� ������ ���������, 5(2), 2017. doi:10.1515/jci-2016-0027

    work page doi:10.1515/jci-2016-0027 2017

  8. [8]

    Bayesian shrinkage priors for penalized synthetic control estimators in the presence of spillovers.����������, 82(2):ujag054, 06 2026

    Esteban Fernández-Morales, Arman Oganisian, and Youjin Lee. Bayesian shrinkage priors for penalized synthetic control estimators in the presence of spillovers.����������, 82(2):ujag054, 06 2026. 10.1093/biomtc/ujag054

    work page doi:10.1093/biomtc/ujag054 2026

Show all 46 references
  1. [9]

    Jennifer L. Hill. Bayesian nonparametric modeling for causal inference.������� �� ������������� ��� ��������� ����������, 20(1):217–240, 2011. 10.1198/jcgs.2010.08162

    2011 arXiv

  2. [10]

    Paul W. Holland. Statistics and causal inference.������� �� ��� �������� ����������� �����������, 81(396):945–960, 1986

    1986

  3. [11]

    Cambridge university press, 2015

    Guido W Imbens and Donald B Rubin.������ ��������� �� ����������� ������� ��� ���������� ��������. Cambridge university press, 2015

    2015

  4. [12]

    Gibbs sampling methods for stick-breaking priors.������� �� ��� �������� ����������� �����������, 96(453):161–173, 2001

    Hemant Ishwaran and Lancelot F James. Gibbs sampling methods for stick-breaking priors.������� �� ��� �������� ����������� �����������, 96(453):161–173, 2001. 10.1198/016214501750332758

    2001 doi

  5. [13]

    A bayesian approach to the g-formula.����������� ������� �� ������� ��������, 27(10):3183–3204, 2018

    Alexander P Keil, Eric J Daza, Stephanie M Engel, Jessie P Buckley, and Jessie K Edwards. A bayesian approach to the g-formula.����������� ������� �� ������� ��������, 27(10):3183–3204, 2018

    2018

  6. [14]

    Randomization-based, bayesian inference of causal effects.������� �� ������ ���������, 11(1): 20220025, 2023

    Thomas Leavitt. Randomization-based, bayesian inference of causal effects.������� �� ������ ���������, 11(1): 20220025, 2023. doi:10.1515/jci-2022-0025

    2023 doi

  7. [15]

    Bayesian causal inference: a critical review.������������� ������ ������� �� ��� ����� ������� �� ������������� �������� ��� ����������� ��������, 381(2247):20220153, 2023

    Fan Li, Peng Ding, and Fabrizia Mealli. Bayesian causal inference: a critical review.������������� ������ ������� �� ��� ����� ������� �� ������������� �������� ��� ����������� ��������, 381(2247):20220153, 2023. 10.1098/rsta.2022.0153

    2023

  8. [16]

    Antonio R. Linero. Simulation-based estimators of analytically intractable causal effects.����������, 78(3):1001– 1017, 05 2021. 10.1111/biom.13499. URL https://doi.org/10.1111/biom.13499

    2021 doi

  9. [17]

    Antonio R. Linero. In nonparametric and high-dimensional models, bayesian ignorability is an informative prior.����� ��� �� ��� �������� ����������� �����������, 12 2023. 10.1080/01621459.2023.2278202

    2023

  10. [18]

    Linero and Joseph L

    Antonio R. Linero and Joseph L. Antonelli. The how and why of bayesian nonparametric causal inference.����� ������������� ����������, 15(1), 2023. https://doi.org/10.1002/wics.1583

    2023 doi

  11. [19]

    Manski and John V

    Charles F. Manski and John V. Pepper. How do right-to-carry laws affect crime rates? coping with ambiguity using bounded-variation assumptions.��� ������ �� ��������� ��� ����������, 100(2):pp. 232–244, 2018. ISSN 00346535, 15309142

    2018

  12. [20]

    On the causal interpretation of randomised interventional indirect effects.������� �� ��� ����� ������� ����� ������� ������ �� ����������� �����������, 85(4):1154–1172, 06 2023

    Caleb H Miles. On the causal interpretation of randomised interventional indirect effects.������� �� ��� ����� ������� ����� ������� ������ �� ����������� �����������, 85(4):1154–1172, 06 2023. 10.1093/jrsssb/qkad066

    2023 doi

  13. [21]

    CRC Press, 2014

    Geert Molenberghs, Garrett Fitzmaurice, Michael G Kenward, Anastasios Tsiatis, and Geert Verbeke.�������� �� ������� ���� �����������. CRC Press, 2014

    2014

  14. [22]

    Priors and propensity scores in bayesian causal inference.������������� �������, 11(1):47, 2025

    Arman Oganisian and Antonio Linero. Priors and propensity scores in bayesian causal inference.������������� �������, 11(1):47, 2025. 10.1353/obs.2025.a956841

    2025 doi

  15. [23]

    Arman Oganisian and Jason A. Roy. A practical introduction to bayesian estimation of causal effects: Parametric and nonparametric approaches.���������� �� ��������, 40:518–551, 1 2021. 10.1002/SIM.8761

    2021 doi

  16. [24]

    Portela, Jeffrey L

    Gerard T. Portela, Jeffrey L. Carson, Sonja A. Swanson, John H. Alexander, Paul C. Hébert, Shaun G. Goodman, Philippe Gabriel Steg, and Others. Effect of four hemoglobin transfusion threshold strategies in patients with acute myocardial infarction and anemia.������ �� ��������...

    2024 doi

  17. [25]

    The dependent dirichlet process and related models.����������� �������, 37(1):24–41, 2022

    Fernando A Quintana, Peter Müller, Alejandro Jara, and Steven N MacEachern. The dependent dirichlet process and related models.����������� �������, 37(1):24–41, 2022. 10.1214/20-STS819. Arman Oganisian, Untangling Sample and Population Level Estimands 21

    2022 doi

  18. [26]

    Thomas S Richardson and James M Robins. Single world intervention graphs (swigs): A unification of the counterfac- tual and graphical approaches to causality.������ ��� ��� ���������� ��� ��� ������ ��������� ���������� �� ���������� ������� ������� �����, 128(30):2013, 2013

    2013

  19. [27]

    James Robins. A new approach to causal inference in mortality studies with a sustained exposure pe- riod—application to control of the healthy worker survivor effect.������������ ���������, 7(9):1393–1512, 1986. https://doi.org/10.1016/0270-0255(86)90088-6

    1986 doi

  20. [28]

    Estimating causal effects of treatments in randomized and nonrandomized studies.������� �� ����������� ����������, 66(5):688, 1974

    Donald B Rubin. Estimating causal effects of treatments in randomized and nonrandomized studies.������� �� ����������� ����������, 66(5):688, 1974. 10.1037/h0037350

    1974 doi

  21. [29]

    Donald B. Rubin. Inference and missing data.����������, 63:581, 12 1976. 10.2307/2335739

    1976 doi

  22. [30]

    Donald B. Rubin. Bayesian inference for causal effects: The role of randomization.������ �� ����������, 6:34–58, 1

  23. [31]

    Donald B. Rubin. The Bayesian Bootstrap.��� ������ �� ����������, 9(1):130 – 134, 1981. 10.1214/aos/1176345338. URL https://doi.org/10.1214/aos/1176345338

    1981

  24. [32]

    Donald B. Rubin. 2 statistical inference for causal effects, with emphasis on applications in epidemiology and med- ical statistics. In C.R. Rao, J.P. Miller, and D.C. Rao, editors,������������ ��� ������� ����������, volume 27 of�������� �� ����������, pages 28–63. Elsevier, ...

    2007 doi

  25. [33]

    On the two-phase framework for joint model and design-based inference.��� ������ �� ����������, 33(6):2789 – 2810, 2005

    Susana Rubin-Bleuer and Ioana Schiopu Kratina. On the two-phase framework for joint model and design-based inference.��� ������ �� ����������, 33(6):2789 – 2810, 2005. 10.1214/009053605000000651. URL https://doi.org/ 10.1214/009053605000000651

    2005 doi

  26. [34]

    Schochet

    Peter Z. Schochet. Estimators for clustered education rcts using the neyman model for causal inference.������� �� ����������� ��� ���������� ����������, 38(3):219–238, 2013. ISSN 10769986, 19351054. URL http://www.jstor.org/ stable/41999423

    2013

  27. [35]

    ignorable

    Chi Wang, Giovanni Parmigiani, and Francesca Dominici. Bayesian effect estimation accounting for adjustment uncer- tainty.����������, 68(3):661–671, 02 2012. 10.1111/j.1541-0420.2011.01731.x. Appendix A Posterior Inference for Sample-Level Estimands In this appendix, we take t...

    2012

  28. [37]

    Given previous parameter draws,ϕ (t) Y , update each subject’s missing potential outcome. For a treated unit, y(t) i (0)|ϕ (t) Y ,yi,li∼∝f(y i,yi(0)|l i;ϕ (t) Y ) and for an untreated unit, y(t) i (1)|ϕ (t) Y ,yi,li∼∝f(y i(1),yi|l i;ϕ (t) Y ) These updates may or may not requi...

  29. [38]

    Combining the imputation of the missing counterfactuals,YM,(t), with the observed dataDO, update the unknownparameters ϕ(t) Y |Y M,(t)∼∝f(ϕ Y ) ∏ i|ai=1 f(yi,y (t) i (0)|l i;ϕY ) ∏ i|ai=0 f(y (t) i (1),yi|l i;ϕY ) Again, in general this may require a MH update as the distribut...

  30. [39]

    the fundamental problem of causal inference

    Compute a posterior draw of the SATE θ(t) = 1 n (∑ i:ai=1 yi−y (t) i (0) + ∑ i:ai=0 y(t) i (1)−y i ) Note that we need not simulate the factual potential outcome - this is observed and fixeda posteriori since it is inDO. Only the counterfactual is simulated since it is unknown...

  31. [40]

    Bayesian g-computation

    and Section 3.3 of [23]. Now that we have posterior drawsΨ(t), we often summarize them via point and interval quantities. For example, a posterior mean forΨcan be approximated as E[Ψ|D O] = ∫ ∫ ∫ Ψ(ϕY1,ϕY0,ϕL)f(ϕY1,ϕY0,ϕL|D O)dϕY1dϕY0dϕL≈ 1 T T∑ t=1 Ψ(t) A 95% credible interva...

  32. [41]

    posterior predictive draws

    For each subjectiin the observed data, simulate potential outcome under treatmenta= 1anda= 0, y(t) i (1)∼f(y(1)|l i;ϕ (t) Y1 ) y(t) i (0)∼f(y(0)|l i;ϕ (t) Y0 ) these are sometimes referred to as “posterior predictive draws.”

  33. [42]

    It departs from the procedure described in Section B in two important ways

    Average the differences to get ˜Ψ(t) = 1 n n∑ i=1 y(t) i (1)−y (t) i (0) ˜Ψ(t) is often taken to be a posterior draw ofΨ, but this would be incorrect. It departs from the procedure described in Section B in two important ways

  34. [43]

    First, rather than simulating covariatesL(1),L (2),...,L (S)∼f(l;ϕ (t) L ), it evaluates the causal effect at eachobservedcovariate value. That is, implicitly, it assumes the covariate distribution is the probability mass function that places massϕ(t) Li on observed covariate ...

  35. [44]

    the causal effect

    Second, rather than simulating largeBvaluesY (1)(a),Y (2)(a),...,Y (B)(a)∼f(y(a)|L=L i;ϕ (t) Ya ) for eachLi, this approach essentially setsB= 1when approximating the CATE at eachli. Thefirstdeparturemeansthatposterioruncertainty(reflectedasvariationacrosstin ˜Ψ(t))doesnotacco...

  36. [45]

    Simulate potential outcome under treatmenta= 1anda= 0, y(t) i (1)∼f(y(1)|l i;ϕ (t) Y1 ) y(t) i (0)∼f(y(0)|l i;ϕ (t) Y0 )

  37. [46]

    for subjecti

    Compute the difference ˜θ(t) i =y (t) i (1)−y (t) i (0) Then, ˜θ(t) i is taken to be a draw of the causal effect “for subjecti”. Across drawst= 1,2,...,T, this is believed to yield a set of posterior draws of this effect. If by “for subjecti”, the analyst means that ˜θ(t) i is...

  38. [1978]

    10.1214/AOS/1176344064

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.