Pith. sign in

REVIEW 2 major objections 1 minor 23 references

Posterior consistency of P\'olya trees for deconvolution under the linear model

T0 review · 2 major / 1 minor · reviewed 2026-06-27 · grok-4.3

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

Pith's one-line read A Pólya tree prior on an unknown density yields posterior concentration in sup-norm for deconvolution from linear noisy observations.

desk verdict The paper extends Castillo 2017 to claim sup-norm posterior consistency for Pólya trees under linear deconvolution, but the abstract leaves the key technical transfer unshown. read the letter →

arxiv 2606.11406 v1 pith:WTZAFKUI submitted 2026-06-09 math.ST stat.TH

Nakul Shenoy , Asaf Weinstein This is my paper · ORCID
classification math.STstat.TH
keywords Pólyatreeposteriorconsistencydeconvolutionlinearmodelsup-normBayesiannonparametricdensityestimation
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

The paper proves that when data follow the linear model Y = X β + ε with the unobserved coefficients βj drawn i.i.d. from an unknown density g0, the posterior induced by a Pólya tree prior on g0 concentrates around the true g0 in the supremum norm. This holds under a condition that the minimum eigenvalue of the design matrix X^T X is bounded away from zero. The argument extends an existing consistency theorem for Pólya trees under direct density estimation to the present deconvolution setting by handling the linear mixing and additive noise. A sympathetic reader would view the result as justification that this Bayesian nonparametric procedure recovers the underlying density reliably from indirect observations as the number of data points grows.

What carries the argument

The Pólya tree prior Π placed on the unknown density g0, with the posterior Π(· | Y) shown to contract around g0 in sup-norm by extending direct-observation consistency arguments to the linear deconvolution setting.

What would settle it

A concrete counterexample in which, for a design matrix satisfying the eigenvalue condition and a fixed g0, the posterior mass outside every sup-norm ball around g0 stays bounded away from zero as the number of rows of X tends to infinity.

Watch

Extended reading notes

Core claim

The central claim is that, for any fixed unknown density g0, the posterior Π(· | Y) from the Pólya tree prior concentrates around g0 in the sup-norm, provided the minimum eigenvalue of X^T X satisfies a suitable condition. The proof adapts results from direct density estimation to account for the structure of the linear model Y = X β + ε where the βj are i.i.d. draws from g0.

Load-bearing premise

The minimum eigenvalue condition on X^T X holds for the given design matrix and the extension of the earlier direct-estimation results introduces no additional gaps.

Editorial extensions

If this is right

  • The posterior can be used to produce consistent estimates of g0 under indirect linear observations.
  • Sup-norm consistency gives uniform control over the entire density rather than weaker integrated distances.
  • The result applies to any fixed g0 without requiring it to belong to a parametric family.
  • The eigenvalue condition on the design matrix is necessary to ensure the linear observations retain enough information about the density.

Reading between the lines

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

  • Credible sets derived from the posterior may attain frequentist coverage guarantees for g0 in large samples.
  • Analogous consistency arguments could apply to other nonparametric priors in similar linear inverse problems.
  • The result suggests studying contraction rates or adaptation properties for this deconvolution model.
  • Practical implementations could test whether the theoretical consistency translates to finite-sample performance under typical design matrices.
Share X Bluesky LinkedIn Reddit HN

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 / 1 minor

Summary. The manuscript claims that a Pólya tree prior on the unknown density g0 yields posterior concentration in the sup-norm around the true g0 when observations follow the linear deconvolution model Y = Xβ + ε with βj i.i.d. from g0. The result requires a minimum-eigenvalue condition on X^T X and is obtained by extending the direct-observation sup-norm consistency theorem of Castillo (2017) to the deconvolution likelihood.

Significance. If the extension is rigorous, the result supplies the first sup-norm posterior consistency guarantee for Pólya trees in a linear deconvolution setting. This would justify the use of the Weinstein et al. (2025) procedure in random-effects and empirical-Bayes applications where only noisy linear combinations of the latent coefficients are observed.

major comments (2)
  1. [Main theorem and proof outline] The central claim rests on the assertion that the local prior-mass, entropy, and sieve arguments of Castillo (2017) transfer to the deconvolution likelihood once a minimum-eigenvalue condition on X^T X is imposed. No explicit verification is supplied that the linear transformation preserves these conditions or that the eigenvalue bound absorbs the additional smoothing; this verification is load-bearing for the sup-norm result.
  2. [Assumptions and main result] The precise statement of the 'suitable condition on the minimum eigenvalue of X^T X' is not given in a form that permits direct checking of the contraction rate or the control of the effective regularity induced by X. Without this, it is impossible to confirm that the deconvolution posterior contracts at the same rate as the direct-observation posterior.
minor comments (1)
  1. The abstract and introduction would benefit from a one-sentence statement of the exact eigenvalue condition and the resulting contraction rate.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. The comments correctly identify that the extension from Castillo (2017) requires more explicit verification of the conditions. We will revise the manuscript to include these details.

read point-by-point responses

  1. Referee: [Main theorem and proof outline] The central claim rests on the assertion that the local prior-mass, entropy, and sieve arguments of Castillo (2017) transfer to the deconvolution likelihood once a minimum-eigenvalue condition on X^T X is imposed. No explicit verification is supplied that the linear transformation preserves these conditions or that the eigenvalue bound absorbs the additional smoothing; this verification is load-bearing for the sup-norm result.

    Authors: We agree with this assessment. The current manuscript states that the arguments transfer under the eigenvalue condition but does not provide the step-by-step verification. In the revised version, we will insert a new subsection (likely Section 3.2 or an appendix) that explicitly verifies how the min-eigenvalue condition on X^T X ensures the local prior mass condition, controls the entropy of the sieve, and absorbs the smoothing effect of the linear model, thereby preserving the sup-norm contraction rate from the direct case. revision: yes

  2. Referee: [Assumptions and main result] The precise statement of the 'suitable condition on the minimum eigenvalue of X^T X' is not given in a form that permits direct checking of the contraction rate or the control of the effective regularity induced by X. Without this, it is impossible to confirm that the deconvolution posterior contracts at the same rate as the direct-observation posterior.

    Authors: We acknowledge that the condition is described qualitatively rather than with explicit quantitative bounds. We will revise the statement of the main theorem to include a precise formulation of the minimum eigenvalue condition, specifying how it scales with the sample size, dimension, and the regularity parameters to ensure the contraction rate remains identical to that in Castillo (2017). This will make it possible to check the rate directly. revision: yes

Circularity Check

0 steps flagged · score 2.0 of 10

Minor self-citation to prior method proposal; main consistency result extends external Castillo (2017) without internal reduction

full rationale

The paper's central claim is a new posterior consistency theorem in sup-norm for the deconvolution setting. It explicitly builds the analysis on Castillo (2017) (external, non-overlapping authors) for the direct-observation case and states that it extends those arguments under an eigenvalue condition on X⊤X. The only self-reference is to Weinstein et al. (2025) for proposing the Pólya tree method itself; this citation is not load-bearing for the consistency proof. No equations reduce by construction, no fitted inputs are relabeled as predictions, and no uniqueness theorem is imported from the authors' own prior work. The derivation is therefore self-contained against external benchmarks once the Castillo extension is granted.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

The paper relies on background results from Castillo (2017) for Pólya tree properties and standard assumptions on the linear model and prior; no free parameters or invented entities are introduced in the abstract.

assumptions (2)
  • domain assumption Pólya tree prior properties and posterior consistency results established in Castillo (2017) for direct density estimation extend to the deconvolution case.
    The analysis builds on and extends results from Castillo (2017).
  • domain assumption The design matrix satisfies a minimum eigenvalue condition sufficient for identifiability in the linear model.
    Required for the main consistency result.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Posterior consistency of P\'olya trees for deconvolution under the linear model." pith.science (2026). https://pith.science/paper/WTZAFKUI

@misc{pith2026260611406,
  author       = {Pith},
  title        = {Pith review of: Posterior consistency of P\'olya trees for deconvolution under the linear model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WTZAFKUI}},
  note         = {Machine review of arXiv:2606.11406}
}
abstract

Several recent works have addressed the problem of deconvolution under a linear model, where the goal is to estimate a completely unknown $G_0$ from a vector of noisy observations $\boldsymbol{Y} = X\boldsymbol{\beta} + \boldsymbol{\epsilon}$, assuming the coefficients $\beta_j$ are i.i.d. unobserved realizations from $G_0$. Assuming $G_0$ has a density $g_0$, we study theoretically a Bayesian nonparametric method proposed in Weinstein et al. (2025) that postulates a P\'olya tree prior $\Pi$ on $g_0$ and bases a deconvolution estimate on the posterior distribution $\Pi(\cdot|\boldsymbol{Y})$. Our main result asserts that under the true model (fixed and unknown $g_0$), and under a suitable condition on the minimum eigenvalue of $X^\top X$, the posterior $\Pi(\cdot|\boldsymbol{Y})$ concentrates around $g_0$ in sup-norm. The analysis presented builds on and extends results from Castillo (2017), where posterior consistency of P\'olya trees was proved for density estimation, the simpler problem of estimating $g_0$ when observing the coefficients $\beta_j$ directly.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 3 canonical work pages

  1. [1]

    C. E. Antoniak. Mixtures of dirichlet processes with applications to bayesian nonparametric problems. The annals of statistics, pages 1152--1174, 1974

    1974

  2. [2]

    D. A. Berry and R. Christensen. Empirical bayes estimation of a binomial parameter via mixtures of dirichlet processes. The Annals of Statistics, 7 0 (3): 0 558--568, 1979

    1979

  3. [3]

    Carbonetto and M

    P. Carbonetto and M. Stephens. Scalable variational inference for bayesian variable selection in regression, and its accuracy in genetic association studies. 2012

    2012

  4. [4]

    Castillo

    I. Castillo. P \'o lya tree posterior distributions on densities. 2017

    2017

  5. [5]

    B. Efron. Empirical bayes deconvolution estimates. Biometrika, 103 0 (1): 0 1--20, 2016

    2016

  6. [6]

    Z. Fan, L. Guan, Y. Shen, and Y. Wu. Gradient flows for empirical bayes in high-dimensional linear models. arXiv preprint arXiv:2312.12708, 2023

    work page Pith review arXiv 2023

  7. [7]

    T. S. Ferguson. Prior distributions on spaces of probability measures. The annals of statistics, pages 615--629, 1974

    1974

  8. [8]

    T. S. Ferguson. Bayesian density estimation by mixtures of normal distributions. In Recent advances in statistics, pages 287--302. Elsevier, 1983

    1983

Show all 23 references
  1. [9]

    Greenshtein

    E. Greenshtein. Consistent empirical bayes estimation of the mean of a mixing distribution with applications to treatment of nonresponse. arXiv preprint arXiv:2511.15373, 2025

    2025

  2. [10]

    Kiefer and J

    J. Kiefer and J. Wolfowitz. Consistency of the maximum likelihood estimator in the presence of infinitely many incidental parameters. The Annals of Mathematical Statistics, pages 887--906, 1956

    1956

  3. [11]

    Y. Kim, W. Wang, P. Carbonetto, and M. Stephens. A flexible empirical bayes approach to multiple linear regression and connections with penalized regression. Journal of Machine Learning Research, 25 0 (185): 0 1--59, 2024

    2024

  4. [12]

    Koenker and I

    R. Koenker and I. Mizera. Convex optimization, shape constraints, compound decisions, and empirical bayes rules. Journal of the American Statistical Association, 109 0 (506): 0 674--685, 2014

    2014

  5. [13]

    M. Lavine. More aspects of polya tree distributions for statistical modelling. The Annals of Statistics, 22 0 (3): 0 1161--1176, 1994

    1994

  6. [14]

    B. G. Lindsay. The geometry of mixture likelihoods: a general theory. The annals of statistics, pages 86--94, 1983

    1983

  7. [15]

    B. G. Lindsay. Mixture models: theory, geometry, and applications. Ims, 1995

    1995

  8. [16]

    A. Y. Lo. On a class of bayesian nonparametric estimates: I. density estimates. The annals of statistics, pages 351--357, 1984

    1984

  9. [17]

    R. D. Mauldin, W. D. Sudderth, and S. C. Williams. P \'o lya trees and random distributions. The Annals of Statistics, pages 1203--1221, 1992

    1992

  10. [18]

    Mukherjee, B

    S. Mukherjee, B. Sen, and S. Sen. A mean field approach to empirical bayes estimation in high-dimensional linear regression. arXiv preprint arXiv:2309.16843, 2023

    2023

  11. [19]

    M. A. Newton. On a nonparametric recursive estimator of the mixing distribution. Sankhy \=a : The Indian Journal of Statistics, Series A , pages 306--322, 2002

    2002

  12. [20]

    Rousseau and C

    J. Rousseau and C. Scricciolo. Wasserstein convergence in bayesian and frequentist deconvolution models. The Annals of Statistics, 52 0 (4): 0 1691--1715, 2024

    2024

  13. [21]

    Vershynin

    R. Vershynin. Introduction to the non-asymptotic analysis of random matrices. In Compressed sensing, pages 210--268. Cambridge Univ. Press, Cambridge, 2012

    2012

  14. [22]

    Y. Wang. On fast computation of the non-parametric maximum likelihood estimate of a mixing distribution. Journal of the Royal Statistical Society Series B: Statistical Methodology, 69 0 (2): 0 185--198, 2007

    2007

  15. [23]

    Weinstein, J

    A. Weinstein, J. Wallin, D. Yekutieli, and M. Bogdan. Nonparametric shrinkage estimation in high dimensional glms via polya trees. Statistica Sinica, 38 0 (4), 2025

    2025

Pith tools

Reviewed June 27, 2026 · model on record in the stance chip above.