pith. sign in

arxiv: 2606.22930 · v1 · pith:KHI4HMNGnew · submitted 2026-06-22 · 📊 stat.CO · astro-ph.IM

First analytical coverage bounds of a fully specified nested sampling algorithm

Johannes Buchner This is my paper

Pith reviewed 2026-06-26 06:22 UTC · model grok-4.3

classification 📊 stat.CO astro-ph.IM
keywords nested samplingMLFriendscoverage boundsbinomial point processmarginal likelihood estimationMonte Carlo samplingBayesian computation
0
0 comments X

The pith

Under a binomial point process model, MLFriends leaves an expected uncovered prior fraction decaying as (1/3 Km)^{-3/2} with negligible bias in the marginal likelihood.

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

The paper derives the first analytical heuristic bounds for coverage in a fully specified nested sampling algorithm called MLFriends. Modeling live points as a homogeneous binomial point process, it shows the expected fraction of the likelihood-restricted prior not covered by the proposal region decays as (1/3 Km)^{-3/2}. For typical numbers of live points K and bootstrap rounds m this fraction is tiny. The resulting bias to the marginal likelihood estimate stays much smaller than the statistical variance from the sampling itself. This matters because it justifies using the algorithm without assuming perfect or asymptotic sampling behavior.

Core claim

MLFriends constructs a proposal region by bootstrap aggregation over the current live points. Under a homogeneous Binomial point process model for these points, the expected uncovered fraction of the likelihood-restricted prior decays as (1/3 Km)^{-3/2} with m bootstrap rounds. This coverage is sufficient that the bias in the marginal likelihood estimate is negligible compared to the inherent statistical variance of a nested sampling run.

What carries the argument

The MLFriends proposal region from bootstrap aggregation over live points, with coverage analyzed under homogeneous Binomial point process model for the live points.

If this is right

  • The bias from incomplete coverage becomes negligible for practical parameter choices.
  • Marginal likelihood estimates from MLFriends nested sampling have bias much smaller than variance.
  • The algorithm provides reliable sampling without needing idealized assumptions.
  • Coverage quality improves with the number of live points and bootstrap rounds following the derived power law.

Where Pith is reading between the lines

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

  • Similar point process models might yield coverage bounds for other nested sampling proposal methods.
  • The heuristic bounds could guide selection of bootstrap rounds m to achieve target coverage.
  • If live point distributions deviate from the binomial model in certain problems, the actual coverage may differ.
  • A rigorous non-heuristic proof would strengthen the result but the current analysis already supports practical use.

Load-bearing premise

The live points can be modeled as a homogeneous Binomial point process.

What would settle it

Numerical experiments measuring the actual uncovered fraction for different K and m and checking if it follows the predicted (1/3 Km)^{-3/2} scaling.

read the original abstract

Nested sampling is a Monte Carlo algorithm for posterior estimation and Bayesian model comparison. It maintains a population of $K$ live points sampled from the prior, and at each iteration discards the lowest-likelihood point and replaces it with a new sample drawn from the prior restricted to exceed the discarded likelihood. Achieving this likelihood-restricted prior sampling efficiently and reliably is the central computational challenge. For low-to-moderate dimensional problems, MLFriends is a general and robust region-based approach that constructs a proposal region by bootstrap aggregation over the current live points and rejects proposals outside this region. We present a self-contained mathematical formulation of MLFriends and derive, under a homogeneous Binomial point process model for the live points, heuristic bounds on the expected fraction of the likelihood-restricted prior not covered by the proposal region. These bounds decay as $(\frac{1}{3}Km)^{-3/2}$, where $m$ is the number of bootstrap rounds, and are negligibly small for practical parameter choices. We show heuristically that the resulting bias in the marginal likelihood estimate is negligible compared to the inherent statistical variance of a nested sampling run. While a fully rigorous treatment remains an open problem, these results provide the first analytical characterisation of a fully specified and practically implementable nested sampling algorithm, without assuming an idealised or asymptotic sampling procedure.

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

2 major / 0 minor

Summary. The manuscript presents a self-contained mathematical formulation of the MLFriends algorithm for nested sampling and derives heuristic bounds on the expected uncovered fraction of the likelihood-restricted prior under a homogeneous Binomial point process model for the live points. These bounds decay as (1/3 Km)^{-3/2} with m bootstrap rounds and the resulting bias in the marginal likelihood estimate is argued heuristically to be negligible compared to the inherent statistical variance; the paper explicitly notes that a fully rigorous treatment remains an open problem.

Significance. If the heuristic derivation is accepted, the work supplies the first analytical characterisation of coverage properties for a fully specified, practically implementable nested sampling algorithm (MLFriends) without idealised or asymptotic assumptions. This is potentially valuable for assessing reliability and guiding parameter selection (K, m) in low-to-moderate dimensional Bayesian computation.

major comments (2)
  1. [Abstract] Abstract: the derivation of the specific decay rate (1/3 Km)^{-3/2} and the negligibility conclusion rests entirely on the homogeneous Binomial point process model for the K live points; this modeling choice is invoked to obtain uniformity and independence, yet the paper provides no justification or diagnostic for why the actual (possibly correlated or non-homogeneous) live-point distribution generated by nested sampling satisfies the assumption.
  2. [Abstract] Abstract: the claim that bias in the marginal likelihood estimate is 'negligibly small' relative to statistical variance is presented as heuristic without an explicit quantitative comparison (e.g., an equation or bound relating the coverage-induced bias term to the usual NS variance expression), which is load-bearing for the practical relevance of the result.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments. We address the two major comments point by point below. Both comments identify areas where additional discussion and explicit comparisons would strengthen the manuscript, and we will revise accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the derivation of the specific decay rate (1/3 Km)^{-3/2} and the negligibility conclusion rests entirely on the homogeneous Binomial point process model for the K live points; this modeling choice is invoked to obtain uniformity and independence, yet the paper provides no justification or diagnostic for why the actual (possibly correlated or non-homogeneous) live-point distribution generated by nested sampling satisfies the assumption.

    Authors: The homogeneous Binomial point process is introduced as a simplifying modeling assumption that permits a self-contained analytical derivation of heuristic coverage bounds under uniformity and independence. This choice enables the explicit decay rate without requiring a full characterization of the (potentially dependent) live-point process generated by the nested sampling iterations. We agree that the manuscript would benefit from explicit discussion of why this model is reasonable for heuristic purposes in low-to-moderate dimensions and from suggested diagnostics. In the revision we will add a paragraph in the discussion section motivating the model, stating its limitations, and outlining possible empirical checks such as comparing predicted versus observed coverage fractions in controlled simulations. revision: yes

  2. Referee: [Abstract] Abstract: the claim that bias in the marginal likelihood estimate is 'negligibly small' relative to statistical variance is presented as heuristic without an explicit quantitative comparison (e.g., an equation or bound relating the coverage-induced bias term to the usual NS variance expression), which is load-bearing for the practical relevance of the result.

    Authors: We accept that the negligibility argument is currently stated heuristically and would be more convincing with an explicit scaling comparison. The manuscript contrasts the coverage error decay with the usual nested-sampling variance term (which scales as O(1/sqrt(K)) or slower with the number of iterations). In the revised version we will insert a short explicit order-of-magnitude relation or bounding argument between the coverage bias term and the standard variance expression, both in the main text and in an updated abstract. revision: yes

Circularity Check

0 steps flagged

No significant circularity; bounds derived from explicit external modeling assumption

full rationale

The paper states its central derivation explicitly proceeds under the modeling assumption that live points form a homogeneous Binomial point process. This assumption is an input to the heuristic bounds on coverage fraction (decaying as (1/3 Km)^{-3/2}) and the subsequent negligibility claim for bias relative to statistical variance. No equations or steps reduce the derived bounds or bias conclusion back to a fitted parameter, self-citation chain, or the target marginal likelihood by construction. The paper acknowledges the heuristic status and that a fully rigorous treatment is open, confirming the derivation does not loop on its own outputs. This is the normal case of a self-contained analysis against an external benchmark model.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The derivation depends on modeling live points as a homogeneous Binomial point process; no free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption live points follow a homogeneous Binomial point process
    Invoked to obtain the coverage bound (abstract section on MLFriends)

pith-pipeline@v0.9.1-grok · 5753 in / 1202 out tokens · 20488 ms · 2026-06-26T06:22:39.789390+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

25 extracted references · 16 canonical work pages · 5 internal anchors

  1. [1]

    Obscuration-dependent evolution of Active Galactic Nuclei

    Hsu, L.-T., Salvato, M., Rangel, C., Aird, J., Merloni, A., Ross, N.: Obscuration- dependent Evolution of Active Galactic Nuclei. ApJ802, 89 (2015) https://doi. org/10.1088/0004-637X/802/2/89 arXiv:1501.02805 [astro-ph.HE]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/802/2/89 2015

  2. [2]

    ArXiv e-prints (2018) arXiv:1807.10677 [astro-ph.HE]

    Baronchelli, L., Nandra, K., Buchner, J.: Relativistic reflection from accretion disks in the population of Active Galactic Nuclei at z=0.5-4. ArXiv e-prints (2018) arXiv:1807.10677 [astro-ph.HE]

    Pith/arXiv arXiv 2018

  3. [3]

    A statistical test for Nested Sampling algorithms

    Buchner, J.: A statistical test for Nested Sampling algorithms. Statistics and Computing, 1–10 (2014) https://doi.org/10.1007/s11222-014-9512-y arXiv:1407.5459 [stat.CO]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/s11222-014-9512-y 2014

  4. [4]

    PASP131(1004), 108005 (2019) https://doi.org/10.1088/1538-3873/ aae7fc

    Buchner, J.: Collaborative Nested Sampling: Big Data versus Complex Physical Models. PASP131(1004), 108005 (2019) https://doi.org/10.1088/1538-3873/ aae7fc

    work page doi:10.1088/1538-3873/ 2019

  5. [5]

    GitHub (2020)

    Buchner, J.: UltraNest v2.2.1. GitHub (2020). https://johannesbuchner.github.io/ UltraNest/

    2020

  6. [6]

    2023, Statistics Surveys, 17, 169, doi: 10.1214/23-SS144

    Buchner, J.: Nested sampling methods. Statistics Surveys17(none), 169–215 (2023) https://doi.org/10.1214/23-SS144 arXiv:2101.09675 [stat.CO]

    work page doi:10.1214/23-ss144 2023

  7. [7]

    Physical Sciences Forum9(1) (2023) https: //doi.org/10.3390/psf2023009017

    Buchner, J.: Snowballing nested sampling. Physical Sciences Forum9(1) (2023) https: //doi.org/10.3390/psf2023009017

    work page doi:10.3390/psf2023009017 2023

  8. [8]

    Chopin, N., Robert, C.P.: Properties of nested sampling. Biometrika (2010) https://doi.org/10.1093/biomet/asq021 http://biomet.oxfordjournals.org/content/early/2010/06/01/biomet.asq021.full.pdf+html Essick,R.,Farr,W.:PrecisionRequirementsforMonteCarloSumswithinHierarchical Bayesian Inference. arXiv e-prints, 2204–00461 (2022) https://doi.org/10.48550/ arX...

    work page doi:10.1093/biomet/asq021 2010

  9. [9]

    Bayesian Statistics8, 491–524 (2007)

    Evans, M.: Discussion of nested sampling for bayesian computations by john skilling. Bayesian Statistics8, 491–524 (2007)

    2007

  10. [10]

    Accuracy Requirements for Empirically-Measured Selection Functions

    Farr, W.M.: Accuracy Requirements for Empirically Measured Selection Functions. 16 Research Notes of the American Astronomical Society3(5), 66 (2019) https://doi. org/10.3847/2515-5172/ab1d5f arXiv:1904.10879 [astro-ph.IM]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/2515-5172/ab1d5f 2019

  11. [11]

    , keywords =

    Feroz, F., Hobson, M.P.: Multimodal nested sampling: an efficient and robust alternative to Markov Chain Monte Carlo methods for astronomical data analyses. MNRAS384, 449–463 (2008) https://doi.org/10.1111/j.1365-2966.2007.12353.x arXiv:0704.3704

    work page doi:10.1111/j.1365-2966.2007.12353.x 2008

  12. [12]

    MNRAS503(1), 1199–1205 (2021) https://doi.org/10.1093/mnras/stab590 arXiv:2010.13884 [stat.CO]

    Fowlie, A., Handley, W., Su, L.: Nested sampling with plateaus. MNRAS503(1), 1199–1205 (2021) https://doi.org/10.1093/mnras/stab590 arXiv:2010.13884 [stat.CO]

    work page doi:10.1093/mnras/stab590 2021

  13. [13]

    nestcheck: diagnostic tests for nested sampling calculations

    Higson, E., Handley, W., Hobson, M., Lasenby, A.: NESTCHECK: diagnostic tests for nested sampling calculations. MNRAS483(2), 2044–2056 (2019) https://doi. org/10.1093/mnras/sty3090 arXiv:1804.06406 [stat.CO]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/sty3090 2044

  14. [14]

    PolyChord: nested sampling for cosmology

    Handley, W.J., Hobson, M.P., Lasenby, A.N.: POLYCHORD: nested sampling for cosmology. MNRAS450, 61–65 (2015) https://doi.org/10.1093/mnrasl/slv047 arXiv:1502.01856 Jasa,T.,Xiang,N.:Nestedsamplingappliedinbayesianroom-acousticsdecayanalysis a. The Journal of the Acoustical Society of America132(5), 3251–3262 (2012)

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnrasl/slv047 2015

  15. [15]

    , keywords =

    Miller, S., Callister, T.A., Farr, W.M.: The Low Effective Spin of Binary Black Holes and Implications for Individual Gravitational-wave Events. ApJ895(2), 128 (2020) https://doi.org/10.3847/1538-4357/ab80c0 arXiv:2001.06051 [astro-ph.HE]

    work page doi:10.3847/1538-4357/ab80c0 2020

  16. [16]

    ApJL638, 51–54 (2006) https://doi.org/10.1086/ 501068 astro-ph/0508461

    Mukherjee, P., Parkinson, D., Liddle, A.R.: A Nested Sampling Algorithm for Cosmological Model Selection. ApJL638, 51–54 (2006) https://doi.org/10.1086/ 501068 astro-ph/0508461

    Pith/arXiv arXiv 2006

  17. [17]

    Bayesian analysis1(4), 833–859 (2006)

    Skilling, J.,et al.: Nested sampling for general bayesian computation. Bayesian analysis1(4), 833–859 (2006)

    2006

  18. [18]

    1467-8624.2007.01075.x

    Shaw, J.R., Bridges, M., Hobson, M.P.: Efficient Bayesian inference for multimodal problems in cosmology. MNRAS378, 1365–1370 (2007) https://doi.org/10.1111/j. 1365-2966.2007.11871.x astro-ph/0701867

    work page doi:10.1111/j 2007

  19. [19]

    AIP Conference Proceedings735(1), 395 (2004) https: //doi.org/10.1063/1.1835238

    Skilling, J.: Nested sampling. AIP Conference Proceedings735(1), 395 (2004) https: //doi.org/10.1063/1.1835238

    work page doi:10.1063/1.1835238 2004

  20. [20]

    Skilling, J.: Nested sampling’s convergence. In: BAYESIAN INFERENCE AND MAXIMUM ENTROPY METHODS IN SCIENCE AND ENGINEERING: The 29th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, vol. 1193, pp. 277–291 (2009). AIP Publishing. http://scitation.aip.org/content/aip/proceeding/aipcp/10.1063/1.3275625 17

    work page doi:10.1063/1.3275625 2009

  21. [21]

    MNRAS493(3), 3132–3158 (2020) https://doi.org/10

    Speagle, J.S.: DYNESTY: a dynamic nested sampling package for estimating Bayesian posteriors and evidences. MNRAS493(3), 3132–3158 (2020) https://doi.org/10. 1093/mnras/staa278 arXiv:1904.02180 [astro-ph.IM]

    Pith/arXiv arXiv 2020

  22. [22]

    arXiv e-prints, 1805–03924 (2018) arXiv:1805.03924 [stat.CO]

    Salomone, R., South, L.F., Drovandi, C.C., Kroese, D.P.: Unbiased and Consistent Nested Sampling via Sequential Monte Carlo. arXiv e-prints, 1805–03924 (2018) arXiv:1805.03924 [stat.CO]

    arXiv 2018

  23. [23]

    arXiv e- prints, 2005–08602 (2020) arXiv:2005.08602 [math.ST]

    Schittenhelm, D., Wacker, P.: Nested Sampling And Likelihood Plateaus. arXiv e- prints, 2005–08602 (2020) arXiv:2005.08602 [math.ST]

    arXiv 2005

  24. [24]

    arXiv e-prints, 2304–06138 (2023) https://doi.org/10.48550/arXiv.2304.06138 arXiv:2304.06138 [astro-ph.IM]

    Talbot, C., Golomb, J.: Growing Pains: Understanding the Impact of Likelihood Uncertainty on Hierarchical Bayesian Inference for Gravitational-Wave Astronomy. arXiv e-prints, 2304–06138 (2023) https://doi.org/10.48550/arXiv.2304.06138 arXiv:2304.06138 [astro-ph.IM]

    work page doi:10.48550/arxiv.2304.06138 2023

  25. [25]

    Walter, C.: Point process-based monte carlo estimation. Statistics and Computing 27(1), 219–236 (2017) arXiv:1412.6368 [cs.CE] Appendix A Implementation details for efficient sampling The following sections present implementation details and efficiency improvements for handling complicated geometries. The metric learning described in Section A.2 determine...

    Pith/arXiv arXiv 2017