pith. sign in

arxiv: 2606.17916 · v1 · pith:H5LGJD22new · submitted 2026-06-16 · 📊 stat.CO · astro-ph.CO· astro-ph.IM· stat.ML

Nested Sampling: A Critical and Comprehensive Theoretical Guide

Luca Martino , Fernando Llorente This is my paper

Pith reviewed 2026-06-26 21:52 UTC · model grok-4.3

classification 📊 stat.CO astro-ph.COastro-ph.IMstat.ML
keywords nested samplingBayesian evidencelikelihood constrained priorsMonte Carlo integrationmarginal likelihoodcosmology applications
0
0 comments X

The pith

Nested sampling computes Bayesian evidence by iteratively sampling from the prior under rising likelihood thresholds.

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

The paper supplies both an introductory tutorial and a critical examination of nested sampling for estimating integrals such as marginal likelihoods. It walks through the complete theoretical derivation, stressing the sequence of constrained prior samplings and the approximations needed to turn those samples into an evidence estimate. The exposition identifies the practical difficulty of drawing from likelihood-restricted priors as the step that governs both the method's efficiency and its potential inaccuracies. By laying out these foundations plainly, the work aims to let new users apply the algorithm correctly and to point experienced users toward targeted improvements.

Core claim

Nested sampling estimates the evidence integral by ordering samples according to increasing likelihood thresholds, replacing the integral with a sum of likelihood times the prior-volume shell between successive thresholds; the derivation relies on treating the constrained prior samples as representative of the prior mass at each threshold, with the main practical requirement being an effective way to generate those constrained draws.

What carries the argument

The likelihood-constrained prior sampling step, which generates points from the prior restricted to likelihood greater than a progressively higher threshold and thereby converts the evidence integral into a one-dimensional quadrature.

If this is right

  • Users can obtain more trustworthy uncertainty estimates once the volume and sampling approximations are tracked explicitly.
  • Variants that replace the constrained sampler with more efficient alternatives become easier to design once the role of that step is isolated.
  • The same framework can be applied to other integrals that can be recast as ordered prior-volume shells.

Where Pith is reading between the lines

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

  • The same constrained-sampling perspective may illuminate why nested sampling sometimes outperforms plain MCMC on multimodal problems.
  • Hybrid schemes that combine nested sampling with gradient-based proposals could be tested by measuring how much they reduce the number of likelihood evaluations needed for a given precision.
  • The review's emphasis on the derivation suggests that existing software implementations could be audited against the exact sequence of approximations described.

Load-bearing premise

That it is possible to generate samples from the prior that are conditioned on the likelihood exceeding a chosen threshold.

What would settle it

A concrete counter-example in which standard nested-sampling code produces systematically biased evidence estimates on a low-dimensional integral whose exact value is known by direct quadrature.

Figures

Figures reproduced from arXiv: 2606.17916 by Fernando Llorente, Luca Martino.

Figure 1
Figure 1. Figure 1: Graphical examples of the two one-dimensional functions [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Two examples of the area below the truncated prior [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Graphical representation of the NS procedure from the [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
read the original abstract

The nested sampling (NS) technique has gained widespread attention, particularly in cosmology and astronomy, due to its ability to efficiently explore high-likelihood regions - a feature akin to an implicit likelihood optimization that underlies its success. While the full theoretical derivation of NS is complex and involves several approximations, the central challenge lies in sampling from the likelihood-constrained priors, which is crucial for its performance. This work provides a comprehensive and detailed exposition of NS derivation, clarifying both its theoretical foundations and practical challenges. We provide a thorough description of the NS procedure, emphasizing both its strengths and potential limitations. In doing so, this work seeks to deepen understanding of the method and to foster the development of future enhancements, novel variants, and more efficient implementations across a wide range of scientific applications. Thus, the main contribution of this work is twofold: it serves both as a tutorial for newcomers to the field and as a critical review for experienced practitioners.

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 / 3 minor

Summary. The manuscript provides a detailed tutorial-style exposition of the nested sampling (NS) algorithm, including its procedure, theoretical foundations (noting that the full derivation involves approximations), and practical challenges, with particular emphasis on sampling from likelihood-constrained priors as the central difficulty. It positions itself as both an introductory guide for newcomers and a critical review for experienced users, primarily in cosmology and astronomy, with the goal of clarifying foundations and fostering future improvements.

Significance. If the exposition accurately and comprehensively covers prior derivations and limitations without introducing unstated gaps, the paper could serve as a useful consolidated reference that aids understanding of NS strengths and weaknesses. The tutorial component may lower barriers for new practitioners, while the critical discussion of sampling challenges could inform variant development; however, as a review rather than a novel contribution, its impact depends on the precision and completeness of the synthesis.

minor comments (3)
  1. Abstract: the claim that NS has an 'implicit likelihood optimization' feature is stated without a supporting reference or derivation sketch; add a brief citation or cross-reference to the relevant section where this is explained.
  2. The manuscript should include an explicit list or table of the specific approximations used in the NS derivation (mentioned in the abstract) with pointers to where each is derived or discussed in the text.
  3. Ensure consistent notation for constrained priors and likelihood thresholds across sections; minor inconsistencies in variable definitions could confuse readers new to the method.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of the manuscript and for recommending minor revision. The work is positioned as both a tutorial for newcomers and a critical review for practitioners, with emphasis on the theoretical foundations of nested sampling and the central practical challenge of sampling from likelihood-constrained priors. We note that no specific major comments were raised in the report.

Circularity Check

0 steps flagged

No significant circularity: review of external method with no new derivations

full rationale

The manuscript is explicitly positioned as a tutorial and critical review of the pre-existing nested sampling algorithm (originally due to Skilling and others). No novel derivation chain is introduced whose steps could reduce to fitted parameters, self-definitions, or self-citations. The exposition describes the standard NS procedure, its approximations, and the known practical difficulty of sampling from likelihood-constrained priors; these descriptions are presented as summaries of established literature rather than as internally generated predictions or uniqueness theorems. Consequently the paper contains no load-bearing steps that collapse by construction to its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a review paper; the abstract introduces no new free parameters, axioms, or invented entities. The exposition rests on the existing literature on nested sampling.

pith-pipeline@v0.9.1-grok · 5688 in / 1014 out tokens · 19741 ms · 2026-06-26T21:52:51.048945+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

34 extracted references · 1 linked inside Pith

  1. [1]

    Ashton, N

    G. Ashton, N. Bernstein, J. Buchner, X. Chen, G. Cs√°nyi, A. Fowlie, F. Feroz, M. Griffiths, W. Handley, M. Habeck, E. Higson, M. Hobson, A. Lasenby, D. Parkinson, L. B. P√°rtay, M. Pitkin, D. Schneider, J. S. Speagle, L. South, J. Veitch, P. Wacker, D. J. Wales, and D. Yallup. Nested sampling for physical scientists.Nature Reviews Methods Primers, 2(1):3...

    2022

  2. [2]

    Betancourt

    M. Betancourt. Nested sampling with constrained Hamiltonian Monte Carlo.AIP Conference Proceedings, 1305(1):165–172, 03 2011

    2011

  3. [3]

    B. J. Brewer, L. B. P´ artay, and G. Cs´ anyi. Diffusive nested sampling.Statistics and Computing, 21(4):649–656, 2011

    2011

  4. [4]

    J. Buchner. Nested sampling methods.Statistics Surveys, 17(none):169 – 215, 2023

    2023

  5. [5]

    M.-H. Chen. Computing marginal likelihoods from a single MCMC output.Statistica Neerlandica, 59(1):16–29, 2005

    2005

  6. [6]

    S. Chib. Marginal likelihood from the Gibbs output.Journal of the American Statistical Association, 90(432):1313–1321, 1995

    1995

  7. [7]

    Chopin and C

    N. Chopin and C. P. Robert. Comments on nested sampling by John Skilling.Bayesian Statistics, 8:491–524, 2007

    2007

  8. [8]

    Chopin and C

    N. Chopin and C. P. Robert. Contemplating evidence: properties, extensions of, and alternatives to nested sampling.Technical Report 2007-46, CEREMADE, Universit Paris Dauphine, pages 1–26, 2007. 26

    2007

  9. [9]

    Chopin and C

    N. Chopin and C. P. Robert. Properties of nested sampling.Biometrika, 97(3):741–755, 2010

    2010

  10. [10]

    Elvira, L

    V. Elvira, L. Martino, and C. P. Robert. Rethinking the effective sample size.International Statistical Review, 90(3):525–550, 2022

    2022

  11. [11]

    Feroz, M

    F. Feroz, M. P. Hobson, and M. Bridges. Multinest: an efficient and robust Bayesian inference tool for cosmology and particle physics.Monthly Notices of the Royal Astronomical Society, 398(4):1601–1614, 2009

    2009

  12. [12]

    Friel and A

    N. Friel and A. N. Pettitt. Marginal likelihood estimation via power posteriors.Journal of the Royal Statistical Society: Series B (Statistical Methodology), 70(3):589–607, 2008

    2008

  13. [13]

    Gelman, J

    A. Gelman, J. B. Carlin, H. S. Stern, D. B. Dunson, A. Vehtari, and D. B. Rubin.Bayesian data analysis. CRC press, 2013

    2013

  14. [14]

    M. Habeck. Nested sampling with demons.AIP Conference Proceedings, 1641(1):121–129, 01 2015

    2015

  15. [15]

    W. J. Handley, M. P. Hobson, and A. N. Lasenby. Polychord: next-generation nested sampling.Monthly Notices of the Royal Astronomical Society, 453(4):4384–4398, 2015

    2015

  16. [16]

    Higson, W

    E. Higson, W. Handley, and A. Hobson, M. land Lasenby. Sampling errors in nested sampling parameter estimation.Bayesian Analysis, 13(3):873–896, 2018

    2018

  17. [17]

    Higson, W

    E. Higson, W. Handley, M. Hobson, and A. Lasenby. Dynamic nested sampling: an improved algorithm for parameter estimation and evidence calculation.Statistics and Computing, 29(5):891–913, 2019

    2019

  18. [18]

    C. R. Keeton. On statistical uncertainty in nested sampling.Monthly Notices of the Royal Astronomical Society, 414(2):1418–1426, 06 2011

    2011

  19. [19]

    K. H. Knuth, M. Habeck, N. K. Malakar, A. M. Mubeen, and B. Placek. Bayesian evidence and model selection.Digital Signal Processing, 47:50–67, 2015. 27

    2015

  20. [20]

    J. S. Liu.Monte Carlo Strategies in Scientific Computing. Springer, 2004

    2004

  21. [21]

    Llorente, L

    F. Llorente, L. Martino, D. Delgado, and J. L´ opez-Santiago. Marginal likelihood computation for model selection and hypothesis testing: An extensive review.SIAM Review, 65(1):3–58, 2023

    2023

  22. [22]

    Martino and V

    L. Martino and V. Elvira. Effective sample size approximations as entropy measures. Computational Statistics, 40(9):5433–5464, Dec 2025

    2025

  23. [23]

    Martino, V

    L. Martino, V. Elvira, and F. Louzada. Effective sample size for importance sampling based on discrepancy measures.Signal Processing, 131:386–401, 2017

    2017

  24. [24]

    Martino, D

    L. Martino, D. Luengo, and J. M´ ıguez. Independent random sampling methods.Springer, 2018

    2018

  25. [25]

    Mikelson and M

    J. Mikelson and M. Khammash. Likelihood-free nested sampling for parameter inference of biochemical reaction networks.PLOS Computational Biology, 16(10):1–24, 10 2020

    2020

  26. [26]

    Mukherjee, D

    P. Mukherjee, D. Parkinson, and A. R. Liddle. A nested sampling algorithm for cosmological model selection.The Astrophysical Journal, 638(2):L51, 2006

    2006

  27. [27]

    R. M. Neal. Slice sampling.The Annals of Statistics, 31(3):705–741, 2003

    2003

  28. [28]

    N. G. Polson and J. G. Scott. Vertical-likelihood Monte Carlo.arXiv preprint arXiv:1409.3601, 2014

    Pith/arXiv arXiv 2014

  29. [29]

    C. P. Robert and G. Casella.Monte Carlo Statistical Methods. Springer, 2004

    2004

  30. [30]

    Skilling

    J. Skilling. Nested sampling.AIP Conference Proceedings, 735(1):395–405, 11 2004

    2004

  31. [31]

    Skilling

    J. Skilling. Nested sampling for general Bayesian computation.Bayesian analysis, 1(4):833– 859, 2006. 28

    2006

  32. [32]

    M. J. Williams, J. Veitch, and C. Messenger. Nested sampling with normalizing flows for gravitational-wave inference.Physical Review D, 103(10):103006, 2021

    2021

  33. [33]

    M. J. Williams, J. Veitch, and C. Messenger. Importance nested sampling with normalising flows.Mach. Learn. Sci. Tech., 4(3):035011, 2023

    2023

  34. [34]

    Zhao and T

    Z. Zhao and T. A. Severini. Integrated likelihood computation methods.Computational Statistics, 32(1):281–313, 2017. 29

    2017