critband: A Python Package for Critical Bandwidth Analysis of Multimodal Distributions

Qihao Wang; Ruiyu Zhang

arxiv: 2605.18686 · v2 · pith:K752YQL2new · submitted 2026-05-18 · 💻 cs.MS · econ.EM

critband: A Python Package for Critical Bandwidth Analysis of Multimodal Distributions

Ruiyu Zhang , Qihao Wang This is my paper

Pith reviewed 2026-05-20 08:24 UTC · model grok-4.3

classification 💻 cs.MS econ.EM

keywords critical bandwidthmultimodal distributionskernel density estimationbimodality detectionPython packagemode countingFFT acceleration

0 comments

The pith

critband supplies a Python implementation of critical bandwidth mode counting that runs several times faster than R equivalents while delivering stable results on separated multimodal data.

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

The paper presents critband as a Python package that fills a gap by bringing Silverman's critical bandwidth approach to multimodal density estimation, complete with FFT acceleration and a bracketed solver for reliable mode counting. It adds practical tools for k-mode detection, component decomposition, bimodality strength measurement, and excess mass calculation. Validation across twelve synthetic benchmarks covering different separations, variances, weights, and sample sizes shows consistent behavior on clearly separated cases and the expected sensitivity near boundaries. Speed tests indicate the package typically finishes each case three to ten times quicker than R's modetest function in the reported setup. Researchers in ecology, economics, genomics, and astronomy would gain from having this capability inside Python workflows rather than switching environments for mode analysis.

Core claim

critband implements critical bandwidth search with a robust bracketed mode-count solver and FFT-accelerated KDE to determine the number of modes in a distribution. The method identifies the smallest bandwidth at which the kernel density estimate exhibits a specified number of modes and extends this to k-mode detection, component decomposition, bimodality strength quantification, and excess mass estimation. Tests on twelve benchmark cases demonstrate stable estimates when modes are clearly separated and the anticipated instability when cases sit near the critical bandwidth boundary.

What carries the argument

critical bandwidth search with a robust bracketed mode-count solver and FFT-accelerated KDE that locates the bandwidth threshold separating different mode counts

If this is right

Users obtain k-mode detection and component decomposition directly from the critical bandwidth procedure.
Bimodality strength and excess mass can be quantified alongside the mode count.
Analysis runs three to ten times faster per case than the comparable R function in the tested configuration.
Expected instability appears for boundary cases near the critical bandwidth, matching the method's design.

Where Pith is reading between the lines

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

The package could serve as a building block for automated multimodal analysis pipelines in Python-based scientific computing.
Direct comparisons on the same real-world datasets across domains would clarify how often boundary instability occurs outside synthetic tests.

Load-bearing premise

The twelve synthetic benchmark cases capture the statistical behavior of real multimodal data encountered in ecology, economics, genomics, and astronomy, particularly near boundary cases.

What would settle it

Applying critband to real datasets from genomics or astronomy with independently verified mode counts and checking whether the reported stability matches the synthetic results would test the claim.

read the original abstract

Multimodal density estimation is a fundamental problem in scientific computing. Determining the number of modes in a distribution is a core numerical challenge with applications across ecology, economics, genomics, and astronomy. While the R ecosystem provides mature tools through the multimode package, the Python ecosystem has lacked an equivalent cohesive implementation. We present critband, a Python package for critical bandwidth bimodality detection based on Silverman's kernel density approach. The package implements critical bandwidth search with a robust bracketed mode-count solver and FFT-accelerated KDE, and provides additional features including k-mode detection, component decomposition, bimodality strength quantification, and excess mass estimation. Validation against twelve benchmark cases spanning separation regimes, unequal variances, unequal weights, and small sample sizes shows stable estimates for clearly separated cases and expected instability for boundary cases. Performance benchmarks show critband is typically 3-10 times faster per case than R's modetest() in the tested setup.

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.

Desk Editor's Note private letter to a colleague

critband is a practical Python port of Silverman's critical bandwidth method with added features and speed benchmarks, but validation stays on synthetic cases.

read the letter

critband is a Python package that implements Silverman's critical bandwidth approach for counting modes in distributions. The main point is that it fills a gap by giving Python users a cohesive tool that was previously only available in R's multimode package, plus extras like k-mode detection, excess mass estimation, and component decomposition. The authors use FFT for faster kernel density estimates and a bracketed solver for mode counting, which are sensible engineering choices. Their tests on twelve synthetic benchmarks cover different separation levels, variances, weights, and small samples, and the outcomes line up with expectations: stable results when modes are clear and more variable near the boundary. They also report that the package runs 3-10 times faster than R's modetest in their setup, which is a concrete practical gain. The soft spot is the exclusive use of synthetic data for validation. Those cases are controlled and useful for checking basic behavior, but real datasets in genomics or astronomy often include heavy tails, outliers, or other features that could shift how the critical bandwidth search performs near decision points. Without any real-data examples, it's not clear how much the reported stability carries over. This work targets applied researchers who need multimodality tools inside Python workflows rather than switching languages. It is not new theory, but the implementation appears careful and the performance claims are testable. I would send it to peer review so referees can examine the code directly and ask for broader testing if needed.

Referee Report

1 major / 2 minor

Summary. The manuscript presents critband, a Python package for critical bandwidth analysis of multimodal distributions based on Silverman's kernel density estimation approach. It implements critical bandwidth search using a robust bracketed mode-count solver and FFT-accelerated KDE, plus extensions for k-mode detection, component decomposition, bimodality strength quantification, and excess mass estimation. Validation on twelve synthetic benchmark cases spanning separation regimes, unequal variances, unequal weights, and small sample sizes reports stable estimates for clearly separated cases and expected instability near boundaries, with runtime benchmarks indicating critband is typically 3-10 times faster than R's modetest() in the tested setup.

Significance. If the implementation is correct and the benchmarks adequately represent the targeted use cases, this package fills a notable gap in the Python ecosystem for multimodal density estimation tools, complementing the mature R multimode package. The performance improvements and additional analytical features could support broader adoption in ecology, economics, genomics, and astronomy, particularly where Python workflows are preferred.

major comments (1)

[Validation section] Validation section (referenced in abstract and described as covering twelve benchmark cases): the central claim of stable estimates for separated cases and expected instability for boundary cases rests entirely on synthetic benchmarks. These may not capture real-data features such as heavy tails, outliers, or non-Gaussian components common in the cited application domains, which could alter the behavior of the Silverman's critical bandwidth search and bracketed solver near decision boundaries. Adding at least one real dataset example from ecology or genomics, or a clear limitations discussion, would strengthen the generalizability of the validation results.

minor comments (2)

[Abstract] Abstract: specify whether the twelve benchmark cases were pre-specified and whether variability across random seeds or multiple runs is reported with error bars.
[Performance benchmarks] Performance benchmarks: provide details on the hardware, Python/R versions, and exact test setup to allow independent reproduction of the 3-10x speedup claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on the validation section. We agree that expanding the discussion of limitations will improve the manuscript and will make the requested revision.

read point-by-point responses

Referee: [Validation section] Validation section (referenced in abstract and described as covering twelve benchmark cases): the central claim of stable estimates for separated cases and expected instability for boundary cases rests entirely on synthetic benchmarks. These may not capture real-data features such as heavy tails, outliers, or non-Gaussian components common in the cited application domains, which could alter the behavior of the Silverman's critical bandwidth search and bracketed solver near decision boundaries. Adding at least one real dataset example from ecology or genomics, or a clear limitations discussion, would strengthen the generalizability of the validation results.

Authors: We agree that the validation relies exclusively on synthetic benchmarks, which, although designed to span separation regimes, unequal variances, unequal weights, and small samples, do not explicitly include heavy tails, outliers, or non-Gaussian components. These features could indeed affect the behavior of the critical bandwidth search and the bracketed mode-count solver near decision boundaries. In the revised manuscript we will add a dedicated limitations subsection that discusses these potential effects, notes the expected increase in instability for boundary cases under such conditions, and clarifies the scope of the current synthetic results. We believe this addition directly addresses the generalizability concern while remaining within the scope of a minor revision. revision: yes

Circularity Check

0 steps flagged

No significant circularity in implementation or validation

full rationale

The paper describes a Python package implementing Silverman's established critical bandwidth method for multimodality detection, along with features like k-mode detection and excess mass estimation. Central claims focus on implementation correctness, stability on synthetic benchmarks, and runtime performance compared to the external R modetest() function. These are evaluated against independent external references and twelve synthetic cases rather than reducing to quantities defined or fitted inside the package. No self-definitional steps, fitted inputs presented as predictions, load-bearing self-citations, or ansatzes smuggled via prior author work appear in the provided content. The work is self-contained against external benchmarks and prior literature.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The package relies on standard kernel density estimation mathematics and Silverman's critical bandwidth definition. No new free parameters are introduced beyond those already present in the reference R implementation. No invented entities are postulated.

axioms (1)

standard math Kernel density estimation with Gaussian kernel produces a valid density estimate whose mode count changes monotonically with bandwidth.
Invoked implicitly when describing the critical bandwidth search; this is a standard property of KDE used in Silverman's original work.

pith-pipeline@v0.9.0 · 5688 in / 1453 out tokens · 39443 ms · 2026-05-20T08:24:07.467051+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The critical bandwidth h_crit(k) is defined as inf{h >0 : ˆf_h has at most k modes}... bracketed binary solver... FFT-accelerated KDE

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

[1]

doi:10.18637/jss.v097.i09. I. K. Baldry et al. Quantifying the bimodal color-magnitude distribution of galaxies.The Astrophysical Journal, 600(2):681–694,

work page doi:10.18637/jss.v097.i09
[2]

doi:10.1086/380092. M. Bessarabova, E. Kirillov, W. Shi, A. Bugrim, Y . Nikolsky, and T. Nikolskaya. Bimodal gene expression patterns in breast cancer.BMC Genomics, 11(Suppl 1):S8,

work page doi:10.1086/380092
[3]

doi:10.1186/1471-2164-11-S1-S8. B. Efron and R. J. Tibshirani.An Introduction to the Bootstrap. Chapman and Hall, New York,

work page doi:10.1186/1471-2164-11-s1-s8
[4]

doi:10.1007/978-1-4899-4541-9. J. Esteban and D. Ray. On the measurement of polarization.Econometrica, 62(4):819–851,

work page doi:10.1007/978-1-4899-4541-9
[5]

doi:10.2307/2951734. N. I. Fisher and J. S. Marron. Mode testing via the excess mass estimate.Biometrika, 88(2):499–517,

work page doi:10.2307/2951734
[6]

doi:10.1093/biomet/88.2.499. P. Hall and M. York. On the calibration of Silverman’s test for multimodality.Statistica Sinica, 11:515–536,

work page doi:10.1093/biomet/88.2.499
[7]

doi:10.1038/s41586-020-2649-

work page doi:10.1038/s41586-020-2649-
[8]

doi:10.1214/aos/1176346577. C. S. Holling. Cross-scale morphology, geometry, and dynamics of ecosystems.Ecological Monographs, 62(4): 447–502,

work page doi:10.1214/aos/1176346577
[9]

doi:10.2307/2937313. M. C. Jones, J. S. Marron, and S. J. Sheather. A brief survey of bandwidth selection for density estimation.Journal of the American Statistical Association, 91(433):401–407,

work page doi:10.2307/2937313
[10]

doi:10.1080/01621459.1996.10476701. D. W. Müller and G. Sawitzki. Excess mass estimates and tests for multimodality.Journal of the American Statistical Association, 86(415):738–746,

work page doi:10.1080/01621459.1996.10476701 1996
[11]

doi:10.1080/01621459.1991.10475103. F. Pedregosa et al. Scikit-learn: Machine learning in Python.Journal of Machine Learning Research, 12:2825– 2830,

work page doi:10.1080/01621459.1991.10475103 1991
[12]

doi:10.1111/j.2517-6161.1981.tb01155.x. B. W. Silverman.Density Estimation for Statistics and Data Analysis. Chapman and Hall/CRC, London,

work page doi:10.1111/j.2517-6161.1981.tb01155.x 1981
[13]

doi:10.1038/s41592-019-0686-2. M. P. Wand and M. C. Jones.Kernel Smoothing. Chapman and Hall, London,

work page doi:10.1038/s41592-019-0686-2

[1] [1]

doi:10.18637/jss.v097.i09. I. K. Baldry et al. Quantifying the bimodal color-magnitude distribution of galaxies.The Astrophysical Journal, 600(2):681–694,

work page doi:10.18637/jss.v097.i09

[2] [2]

doi:10.1086/380092. M. Bessarabova, E. Kirillov, W. Shi, A. Bugrim, Y . Nikolsky, and T. Nikolskaya. Bimodal gene expression patterns in breast cancer.BMC Genomics, 11(Suppl 1):S8,

work page doi:10.1086/380092

[3] [3]

doi:10.1186/1471-2164-11-S1-S8. B. Efron and R. J. Tibshirani.An Introduction to the Bootstrap. Chapman and Hall, New York,

work page doi:10.1186/1471-2164-11-s1-s8

[4] [4]

doi:10.1007/978-1-4899-4541-9. J. Esteban and D. Ray. On the measurement of polarization.Econometrica, 62(4):819–851,

work page doi:10.1007/978-1-4899-4541-9

[5] [5]

doi:10.2307/2951734. N. I. Fisher and J. S. Marron. Mode testing via the excess mass estimate.Biometrika, 88(2):499–517,

work page doi:10.2307/2951734

[6] [6]

doi:10.1093/biomet/88.2.499. P. Hall and M. York. On the calibration of Silverman’s test for multimodality.Statistica Sinica, 11:515–536,

work page doi:10.1093/biomet/88.2.499

[7] [7]

doi:10.1038/s41586-020-2649-

work page doi:10.1038/s41586-020-2649-

[8] [8]

doi:10.1214/aos/1176346577. C. S. Holling. Cross-scale morphology, geometry, and dynamics of ecosystems.Ecological Monographs, 62(4): 447–502,

work page doi:10.1214/aos/1176346577

[9] [9]

doi:10.2307/2937313. M. C. Jones, J. S. Marron, and S. J. Sheather. A brief survey of bandwidth selection for density estimation.Journal of the American Statistical Association, 91(433):401–407,

work page doi:10.2307/2937313

[10] [10]

doi:10.1080/01621459.1996.10476701. D. W. Müller and G. Sawitzki. Excess mass estimates and tests for multimodality.Journal of the American Statistical Association, 86(415):738–746,

work page doi:10.1080/01621459.1996.10476701 1996

[11] [11]

doi:10.1080/01621459.1991.10475103. F. Pedregosa et al. Scikit-learn: Machine learning in Python.Journal of Machine Learning Research, 12:2825– 2830,

work page doi:10.1080/01621459.1991.10475103 1991

[12] [12]

doi:10.1111/j.2517-6161.1981.tb01155.x. B. W. Silverman.Density Estimation for Statistics and Data Analysis. Chapman and Hall/CRC, London,

work page doi:10.1111/j.2517-6161.1981.tb01155.x 1981

[13] [13]

doi:10.1038/s41592-019-0686-2. M. P. Wand and M. C. Jones.Kernel Smoothing. Chapman and Hall, London,

work page doi:10.1038/s41592-019-0686-2