pith. sign in

arxiv: 2606.07947 · v1 · pith:NIZHFNY7new · submitted 2026-06-06 · 📊 stat.ME · math.ST· stat.AP· stat.TH

Bayesian Global Fr\'echet Regression via Weak Conditional Expectations

Simon Fontaine , Bing Li , Lingzhou Xue This is my paper

Pith reviewed 2026-06-27 19:42 UTC · model grok-4.3

classification 📊 stat.ME math.STstat.APstat.TH
keywords Bayesian Fréchet regressionweak conditional expectationsmetric space regressionobject-valued dataprior shrinkagemodel misspecificationmicrobiome compositional data
0
0 comments X

The pith

A Bayesian Fréchet regression framework reduces object-valued problems to scalar regressions via a novel Bayes rule that remains valid under moment conditions.

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

The paper develops a Bayesian method for Fréchet regression, where responses live in general metric spaces while predictors are Euclidean. It defines a Fréchet Bayes rule that converts the regression into a set of ordinary scalar regressions, letting users blend prior information with the data in a controlled way. The construction is first shown under Gaussian assumptions but then proved to hold more generally whenever only moment conditions are met, using weak conditional expectations. Simulations and a microbiome compositional data example illustrate that an auxiliary cohort can supply an informative prior that improves predictions in small targeted studies.

Core claim

Targeting a novel Fréchet Bayes rule reduces the object-valued regression problem to a collection of tractable scalar regression tasks. This rule supports a controlled interpolation between the prior and the data-driven frequentist estimate, enabling shrinkage toward informed values. The framework stays valid under model misspecification provided only moment conditions hold, because weak conditional expectations suffice for the key identities.

What carries the argument

The Fréchet Bayes rule, which converts metric-space regression into scalar tasks, together with weak conditional expectations that secure validity from moment conditions alone.

If this is right

  • Prior information from auxiliary samples can be incorporated into global Fréchet regression for small target studies.
  • Shrinkage toward informed values becomes available for object-valued responses.
  • The reduction to scalar tasks makes the method computationally tractable for nonlinear global regression.
  • Robustness under misspecification broadens applicability beyond correctly specified Gaussian models.

Where Pith is reading between the lines

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

  • Existing Bayesian software for scalar regression could be reused directly after the reduction step.
  • The same moment-based argument might apply to other Fréchet-type problems outside regression.
  • Performance gains from auxiliary priors could be tested in additional domains such as shape or network data.

Load-bearing premise

Weak conditional expectations based on moment conditions are sufficient to keep the Bayes rule valid even when the model is misspecified.

What would settle it

In the microbiome application, predictive performance on the target cohort would fail to improve when the auxiliary-cohort prior is used compared with the frequentist estimate alone.

Figures

Figures reproduced from arXiv: 2606.07947 by Bing Li, Lingzhou Xue, Simon Fontaine.

Figure 1
Figure 1. Figure 1: Bayesian global Fréchet regression for univariate Gaussian distributional responses. [PITH_FULL_IMAGE:figures/full_fig_p023_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Bayesian global Fréchet regression for spherical responses on the unit sphere. Left [PITH_FULL_IMAGE:figures/full_fig_p025_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Prior misspecification and robustness analysis for spherical responses on the unit [PITH_FULL_IMAGE:figures/full_fig_p026_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Cross-study predictive portability for microbiome compositional data across the [PITH_FULL_IMAGE:figures/full_fig_p029_4.png] view at source ↗
read the original abstract

Fr\'echet regression provides a versatile framework for modeling responses in metric spaces with Euclidean predictors, yet current methodologies rely almost exclusively on frequentist approaches. We propose a Bayesian framework for Fr\'echet regression that offers a principled way of incorporating prior information into nonlinear global Fr\'echet regression. By targeting a novel Fr\'echet Bayes rule, we reduce the object-valued regression problem to a collection of tractable scalar regression tasks. Our approach allows for a controlled interpolation between the prior and the data-driven frequentist estimate, facilitating effective shrinkage toward informed values. While initially derived under Gaussian assumptions, we demonstrate that our framework is robust to model misspecification by establishing its validity under moment conditions via weak conditional expectations. The numerical properties of the proposed methodology are demonstrated in simulation studies and an application to microbiome compositional data, where we show that leveraging an auxiliary cohort to inform the prior significantly enhances predictive performance in a targeted, small-scale study

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 / 2 minor

Summary. The paper proposes a Bayesian framework for global Fréchet regression of metric-space responses on Euclidean predictors. It introduces a Fréchet Bayes rule that reduces the object-valued problem to a collection of scalar regression tasks, permitting controlled shrinkage between a prior and the frequentist Fréchet estimator. The method is first derived under Gaussian assumptions and then extended to validity under model misspecification by replacing ordinary conditional expectations with weak conditional expectations defined via moment conditions alone. Numerical performance is illustrated in simulations and a microbiome compositional-data application that uses an auxiliary cohort to inform the prior.

Significance. If the reduction to scalar tasks and the misspecification robustness both hold, the work would supply the first principled Bayesian procedure for global Fréchet regression, with practical value for small-sample prediction problems that can borrow strength from auxiliary data. The explicit interpolation between prior and data-driven estimates is a clear methodological contribution over existing frequentist Fréchet methods.

major comments (2)
  1. [§3.2] §3.2 (Weak conditional expectations and the Fréchet Bayes rule): The argument that the Fréchet mean defined by the weak conditional expectation coincides with the argmin of the integrated squared-distance functional rests on moment conditions alone. In a general metric space the squared-distance map need not be continuous or convex with respect to weak convergence; the manuscript does not supply the additional topological or convexity hypotheses that would justify interchanging the weak limit and the variational definition. Without this step the claimed validity under misspecification does not follow.
  2. [Theorem 3.1] Theorem 3.1 and the subsequent reduction to scalar regressions: The passage from the object-valued posterior Fréchet mean to a collection of independent scalar regressions is asserted after the weak-conditional-expectation construction. The proof sketch does not verify that the metric-space geometry is preserved once the full conditional law is replaced by its weak version; a counter-example in a non-Hilbert metric space would falsify the reduction.
minor comments (2)
  1. The notation for the weak conditional expectation operator is introduced without an explicit comparison to the ordinary conditional expectation; a short remark clarifying when the two coincide would aid readability.
  2. Figure 2 (simulation results) reports point estimates but omits variability bands for the Bayesian versus frequentist estimators; adding these would strengthen the visual comparison.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments on our manuscript. We provide point-by-point responses to the major comments below, and we plan to incorporate clarifications in a revised version.

read point-by-point responses

  1. Referee: [§3.2] §3.2 (Weak conditional expectations and the Fréchet Bayes rule): The argument that the Fréchet mean defined by the weak conditional expectation coincides with the argmin of the integrated squared-distance functional rests on moment conditions alone. In a general metric space the squared-distance map need not be continuous or convex with respect to weak convergence; the manuscript does not supply the additional topological or convexity hypotheses that would justify interchanging the weak limit and the variational definition. Without this step the claimed validity under misspecification does not follow.

    Authors: We appreciate the referee pointing out this technical detail. Our construction of the weak conditional expectation is based on moment conditions, and the Fréchet mean is defined as the minimizer of the expected squared distance. In the paper, we implicitly rely on the properties of the metric space that ensure the necessary continuity for the interchange, as is common in Fréchet regression literature. However, to make this explicit, we will add a remark in Section 3.2 specifying the conditions (e.g., lower semi-continuity of the squared distance functional under weak convergence) under which the result holds. This will strengthen the justification for validity under misspecification. revision: partial

  2. Referee: [Theorem 3.1] Theorem 3.1 and the subsequent reduction to scalar regressions: The passage from the object-valued posterior Fréchet mean to a collection of independent scalar regressions is asserted after the weak-conditional-expectation construction. The proof sketch does not verify that the metric-space geometry is preserved once the full conditional law is replaced by its weak version; a counter-example in a non-Hilbert metric space would falsify the reduction.

    Authors: Regarding the reduction to scalar regressions in Theorem 3.1, the weak conditional expectation is designed such that the Fréchet Bayes rule decomposes the problem into scalar tasks while preserving the essential geometry through the moment conditions. The proof sketch in the manuscript outlines this decomposition. We disagree that a counter-example in a non-Hilbert space would necessarily falsify the reduction, as our framework is developed for general metric spaces where the Fréchet mean exists and the weak version maintains the variational property. Nevertheless, we will expand the proof in the revision to include a verification step showing that the geometry is preserved under the stated assumptions. revision: partial

Circularity Check

0 steps flagged

No circularity: derivation reduces object regression to scalar tasks via independent weak conditional expectation construction.

full rationale

The abstract and reader's summary indicate the Fréchet Bayes rule is a novel targeting construction that converts the metric-space problem into scalar regressions, with robustness shown under moment conditions. No quoted equation or self-citation chain reduces the central claim to a fitted input or prior self-result by definition. The reduction and misspecification validity are presented as derived results rather than tautological renamings or load-bearing self-citations. This is the expected non-finding for a paper whose core steps remain externally verifiable.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review limits visibility into specific parameters or entities; the robustness claim rests on an unelaborated domain assumption about moment conditions and weak conditional expectations.

axioms (1)
  • domain assumption Validity of the Fréchet Bayes rule under moment conditions via weak conditional expectations
    Invoked in abstract to establish robustness to model misspecification beyond Gaussian assumptions.

pith-pipeline@v0.9.1-grok · 5693 in / 1255 out tokens · 19893 ms · 2026-06-27T19:42:43.211666+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

240 extracted references · 150 canonical work pages · 9 internal anchors

  1. [1]

    arXiv preprint arXiv:2506.22754 , year=

    Doubly robust estimation of causal effects for random object outcomes with continuous treatments , author=. arXiv preprint arXiv:2506.22754 , year=

    arXiv

  2. [2]

    Journal of Multivariate Analysis , volume=

    Nonlinear sufficient dimension reduction for distribution-on-distribution regression , author=. Journal of Multivariate Analysis , volume=. 2024 , publisher=

    2024

  3. [3]

    arXiv.org , author =

    Variable. arXiv.org , author =

  4. [4]

    Reproducing kernel

    Berlinet, Alain and Thomas-Agnan, Christine , year =. Reproducing kernel

  5. [5]

    Fréchet sufficient dimension reduction for random objects , volume =

    Ying, Chao and Yu, Zhou , year =. Fréchet sufficient dimension reduction for random objects , volume =. Biometrika , publisher =

  6. [6]

    Causal inference on distribution functions , volume =

    Lin, Zhenhua and Kong, Dehan and Wang, Linbo , year =. Causal inference on distribution functions , volume =. Journal of the Royal Statistical Society Series B , publisher =

  7. [7]

    Nonlinear sufficient dimension reduction for distribution-on-distribution regression , volume =

    Zhang, Qi and Li, Bing and Xue, Lingzhou , year =. Nonlinear sufficient dimension reduction for distribution-on-distribution regression , volume =. Journal of Multivariate Analysis , publisher =

  8. [8]

    Doubly robust estimation of causal effects for random object outcomes with continuous treatments , journal =

    Bhattacharjee, Satarupa and Li, Bing and Wu, Xiao and Xue, Lingzhou , year =. Doubly robust estimation of causal effects for random object outcomes with continuous treatments , journal =

  9. [9]

    Practical

    Snoek, Jasper and Larochelle, Hugo and Adams, Ryan P , year =. Practical. Advances in

  10. [10]

    Automatic relevance determination for neural networks , journal =

    MacKay, David JC and Neal, Radford M , year =. Automatic relevance determination for neural networks , journal =

  11. [11]

    Adaptive

    Villani, Paolo and Unger, Jörg and Weiser, Martin , month = apr, year =. Adaptive. doi:10.48550/arXiv.2404.19459 , abstract =

    work page doi:10.48550/arxiv.2404.19459

  12. [12]

    Journal of Statistical Planning and Inference , author =

    Bayesian variable selection using an adaptive powered correlation prior , volume =. Journal of Statistical Planning and Inference , author =. 2009 , keywords =

    2009

  13. [13]

    Comparison of different approaches of finding the positive definite metric in pseudo-Hermitian theories

    Mulder, Joris , month = jan, year =. Bayesian. The American Statistician , publisher =. doi:10.1080/00031305.2022.2028675 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1080/00031305.2022.2028675 2022

  14. [14]

    On assessing prior distributions and

    Zellner, Arnold , year =. On assessing prior distributions and. Bayesian Inference and Decision Techniques , publisher =

  15. [15]

    , year =

    ZELLNER, A. , year =. On assessing prior distributions and. Bayesian Inference and Decision techniques , publisher =

  16. [16]

    Zellner, Arnold , month = nov, year =. Optimal. The American Statistician , publisher =. doi:10.1080/00031305.1988.10475585 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1080/00031305.1988.10475585 1988

  17. [17]

    Fong, C. H. E. , year =. The predictive view of

  18. [18]

    Dellaporta, Charita and Knoblauch, Jeremias and Damoulas, Theodoros and Briol, Francois-Xavier , month = may, year =. Robust. Proceedings of

  19. [19]

    Bayesian fractional posteriors , volume =

    Bhattacharya, Anirban and Pati, Debdeep and Yang, Yun , month = feb, year =. Bayesian fractional posteriors , volume =. The Annals of Statistics , publisher =. doi:10.1214/18-AOS1712 , abstract =

    work page doi:10.1214/18-aos1712

  20. [20]

    Journal of the Royal Statistical Society Series B , author =

    Martingale posterior distributions , volume =. Journal of the Royal Statistical Society Series B , author =. 2023 , pages =. doi:10.1093/jrsssb/qkad005 , abstract =

    work page doi:10.1093/jrsssb/qkad005 2023

  21. [22]

    Biometrika , author =

    On the marginal likelihood and cross-validation , volume =. Biometrika , author =. 2020 , pages =. doi:10.1093/biomet/asz077 , abstract =

    work page doi:10.1093/biomet/asz077 2020

  22. [23]

    Biometrika , author =

    Assigning a value to a power likelihood in a general. Biometrika , author =. 2017 , pages =. doi:10.1093/biomet/asx010 , abstract =

    work page doi:10.1093/biomet/asx010 2017

  23. [24]

    Journal of the Royal Statistical Society Series B , author =

    Robust. Journal of the Royal Statistical Society Series B , author =. 2022 , pages =. doi:10.1111/rssb.12500 , abstract =

    work page doi:10.1111/rssb.12500 2022

  24. [25]

    Journal of the Royal Statistical Society Series A: Statistics in Society , author =

    Beyond. Journal of the Royal Statistical Society Series A: Statistics in Society , author =. 2017 , pages =. doi:10.1111/rssa.12276 , abstract =

    work page doi:10.1111/rssa.12276 2017

  25. [26]

    Inconsistency of

    Grünwald, Peter and Ommen, Thijs van , month = dec, year =. Inconsistency of. Bayesian Analysis , publisher =. doi:10.1214/17-BA1085 , abstract =

    work page doi:10.1214/17-ba1085

  26. [28]

    Journal of Machine Learning Research , author =

    On the properties of variational approximations of. Journal of Machine Learning Research , author =. 2016 , pages =

    2016

  27. [29]

    Journal of Machine Learning Research , author =

    The. Journal of Machine Learning Research , author =. 2023 , pages =

    2023

  28. [30]

    Journal of Machine Learning Research , author =

    An. Journal of Machine Learning Research , author =. 2022 , pages =

    2022

  29. [31]

    and Drovandi, Christopher and Frazier, David T

    Nott, David J. and Drovandi, Christopher and Frazier, David T. , month = apr, year =. Bayesian. Annual Review of Statistics and Its Application , publisher =. doi:10.1146/annurev-statistics-040522-015915 , abstract =

    work page doi:10.1146/annurev-statistics-040522-015915

  30. [33]

    Farquhar, Greg and Baumli, Kate and Marinho, Zita and Filos, Angelos and Hessel, Matteo and van Hasselt, Hado P and Silver, David , year =. Self-. Advances in

  31. [34]

    Self-consistency: a fundamental concept in statistics , volume =

    Flury, Bernard and Tarpey, Thaddeus , month = sep, year =. Self-consistency: a fundamental concept in statistics , volume =. Statistical Science , publisher =. doi:10.1214/ss/1032280215 , abstract =

    work page doi:10.1214/ss/1032280215

  32. [35]

    Statistical

    Single. Statistical. 2019 , note =. doi:10.1002/9781119482260.ch4 , abstract =

    work page doi:10.1002/9781119482260.ch4 2019

  33. [36]

    Statistical

    Models for. Statistical. 2019 , note =. doi:10.1002/9781119482260.ch12 , abstract =

    work page doi:10.1002/9781119482260.ch12 2019

  34. [37]

    Statistical

    Bayes and. Statistical. 2019 , note =. doi:10.1002/9781119482260.ch10 , abstract =

    work page doi:10.1002/9781119482260.ch10 2019

  35. [38]

    Statistical

    Maximum. Statistical. 2019 , note =. doi:10.1002/9781119482260.ch8 , abstract =

    work page doi:10.1002/9781119482260.ch8 2019

  36. [39]

    Statistical

    Large-. Statistical. 2019 , note =. doi:10.1002/9781119482260.ch9 , abstract =

    work page doi:10.1002/9781119482260.ch9 2019

  37. [40]

    Biometrika , author =

    Maximum likelihood estimation via the. Biometrika , author =. 1993 , pages =. doi:10.1093/biomet/80.2.267 , abstract =

    work page doi:10.1093/biomet/80.2.267 1993

  38. [41]

    A view of the

    Neal, Radford M and Hinton, Geoffrey E , year =. A view of the. Learning in graphical models , publisher =

  39. [42]

    The Annals of Statistics , author =

    Nonlinear global. The Annals of Statistics , author =. 2025 , keywords =. doi:10.1214/24-AOS2457 , number =

    work page doi:10.1214/24-aos2457 2025

  40. [43]

    Food Bioscience , author =

    Comprehensive database for food-gut microbiota-disease interactions (. Food Bioscience , author =. 2024 , keywords =. doi:10.1016/j.fbio.2024.104091 , urldate =

    work page doi:10.1016/j.fbio.2024.104091 2024

  41. [44]

    Gut , author =

    Metagenomic analysis of faecal microbiome as a tool towards targeted non-invasive biomarkers for colorectal cancer , volume =. Gut , author =. 2017 , keywords =. doi:10.1136/gutjnl-2015-309800 , language =

    work page doi:10.1136/gutjnl-2015-309800 2017

  42. [45]

    The Annals of Statistics , author =

    Conformal inference for random objects , volume =. The Annals of Statistics , author =. 2025 , keywords =. doi:10.1214/25-AOS2495 , number =

    work page doi:10.1214/25-aos2495 2025

  43. [46]

    Journal of Computational and Graphical Statistics , author =

    Kernel. Journal of Computational and Graphical Statistics , author =. 2005 , keywords =. doi:10.1198/106186005X25619 , number =

    work page doi:10.1198/106186005x25619 2005

  44. [47]

    Importance

    Xu, Liyuan and Chen, Yutian and Doucet, Arnaud and Gretton, Arthur , month = jun, year =. Importance. Proceedings of the 39th

  45. [48]

    Journal of the American Statistical Association119(547), 2328–2344 (2024) https://doi.org/10.1080/01621459.2023.2257889

    Dimension Reduction for. Journal of the American Statistical Association , author =. 2024 , keywords =. doi:10.1080/01621459.2023.2277406 , number =

    work page doi:10.1080/01621459.2023.2277406 2024

  46. [49]

    , year =

    Yi, Grace Y. , year =. Composite. Wiley

  47. [50]

    Wasserstein

    Xu, Haoshu and Li, Hongzhe , month = apr, year =. Wasserstein. doi:10.48550/arXiv.2404.03878 , urldate =

    work page doi:10.48550/arxiv.2404.03878

  48. [51]

    2021 , pages =

    Nucleic Acids Research , author =. 2021 , pages =. doi:10.1093/nar/gkaa851 , number =

    work page doi:10.1093/nar/gkaa851 2021

  49. [52]

    World Journal of Methodology , author =

    Update on the gut microbiome in health and diseases , volume =. World Journal of Methodology , author =. 2024 , pages =. doi:10.5662/wjm.v14.i1.89196 , number =

    work page doi:10.5662/wjm.v14.i1.89196 2024

  50. [53]

    When does a

    Steinwart, Ingo , month = mar, year =. When does a. doi:10.48550/arXiv.2407.11898 , urldate =

    work page doi:10.48550/arxiv.2407.11898

  51. [54]

    Statistical science : a review journal of the Institute of Mathematical Statistics , author =

    Variable. Statistical science : a review journal of the Institute of Mathematical Statistics , author =. 2011 , pages =. doi:10.1214/11-STS354 , number =

    work page doi:10.1214/11-sts354 2011

  52. [55]

    The Journal of Experimental Medicine , author =

    Squamous cell carcinoma subverts adjacent histologically normal epithelium to promote lateral invasion , volume =. The Journal of Experimental Medicine , author =. 2021 , keywords =. doi:10.1084/jem.20200944 , language =

    work page doi:10.1084/jem.20200944 2021

  53. [56]

    Bioinformatics , author =

    Information theoretic generalized. Bioinformatics , author =. 2020 , pages =. doi:10.1093/bioinformatics/btaa614 , number =

    work page doi:10.1093/bioinformatics/btaa614 2020

  54. [57]

    Wahba, Grace and Gu, Chong and Wang, Yuedong and Chappell, Richard , year =. Soft. The

  55. [58]

    Conditional

    Sugiyama, Masashi and Takeuchi, Ichiro and Suzuki, Taiji and Kanamori, Takafumi and Hachiya, Hirotaka and Okanohara, Daisuke , month = mar, year =. Conditional. Proceedings of the

  56. [59]

    Annals of the Institute of Statistical Mathematics , author =

    Direct importance estimation for covariate shift adaptation , volume =. Annals of the Institute of Statistical Mathematics , author =. 2008 , pages =. doi:10.1007/s10463-008-0197-x , language =

    work page doi:10.1007/s10463-008-0197-x 2008

  57. [60]

    Annals of the Institute of Statistical Mathematics , author =

    Density-ratio matching under the. Annals of the Institute of Statistical Mathematics , author =. 2012 , pages =. doi:10.1007/s10463-011-0343-8 , language =

    work page doi:10.1007/s10463-011-0343-8 2012

  58. [61]

    Hilbert space embeddings of conditional distributions with applications to dynamical systems , isbn =

    Song, Le and Huang, Jonathan and Smola, Alex and Fukumizu, Kenji , month = jun, year =. Hilbert space embeddings of conditional distributions with applications to dynamical systems , isbn =. Proceedings of the 26th. doi:10.1145/1553374.1553497 , urldate =

    work page doi:10.1145/1553374.1553497

  59. [62]

    IEEE Signal Processing Magazine , author =

    Kernel. IEEE Signal Processing Magazine , author =. 2013 , keywords =. doi:10.1109/MSP.2013.2252713 , number =

    work page doi:10.1109/msp.2013.2252713 2013

  60. [63]

    Shimizu, Eiki and Fukumizu, Kenji and Sejdinovic, Dino , month = mar, year =. Neural-. doi:10.48550/arXiv.2403.10859 , urldate =

    work page doi:10.48550/arxiv.2403.10859

  61. [64]

    Hypothesis testing using pairwise distances and associated kernels (with Appendix)

    Sejdinovic, Dino and Gretton, Arthur and Sriperumbudur, Bharath and Fukumizu, Kenji , month = may, year =. Hypothesis testing using pairwise distances and associated kernels (with. doi:10.48550/arXiv.1205.0411 , urldate =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1205.0411

  62. [65]

    Technometrics , author =

    Can’t. Technometrics , author =. 2021 , keywords =. doi:10.1080/00401706.2020.1791254 , number =

    work page doi:10.1080/00401706.2020.1791254 2021

  63. [66]

    Proceedings of the IEEE , author =

    Taking the. Proceedings of the IEEE , author =. 2016 , keywords =. doi:10.1109/JPROC.2015.2494218 , number =

    work page doi:10.1109/jproc.2015.2494218 2016

  64. [67]

    International Journal of Neural Systems , author =

    Gaussian processes for machine learning , volume =. International Journal of Neural Systems , author =. 2004 , keywords =. doi:10.1142/S0129065704001899 , number =

    work page doi:10.1142/s0129065704001899 2004

  65. [68]

    Rossi, Simone and Heinonen, Markus and Bonilla, Edwin and Shen, Zheyang and Filippone, Maurizio , month = mar, year =. Sparse. Proceedings of

  66. [69]

    Variational

    Titsias, Michalis , month = apr, year =. Variational. Proceedings of the

  67. [70]

    Communications in Statistics - Theory and Methods , author =

    Fiducial and. Communications in Statistics - Theory and Methods , author =. 2015 , keywords =. doi:10.1080/03610926.2013.823207 , number =

    work page doi:10.1080/03610926.2013.823207 2015

  68. [71]

    Ventura, Laura and Racugno, Walter , editor =. Pseudo-. Topics on. 2016 , keywords =. doi:10.1007/978-3-319-44093-4_19 , language =

    work page doi:10.1007/978-3-319-44093-4_19 2016

  69. [72]

    Statistica Sinica , author =

    An. Statistica Sinica , author =. 2011 , pages =

    2011

  70. [73]

    Journal of Machine Learning for Modeling and Computing , author =

    A. Journal of Machine Learning for Modeling and Computing , author =. 2020 , keywords =. doi:10.1615/JMachLearnModelComput.2020035155 , number =

    work page doi:10.1615/jmachlearnmodelcomput.2020035155 2020

  71. [74]

    and Fan, Yanan and Beaumont, Mark , month = sep, year =

    Sisson, Scott A. and Fan, Yanan and Beaumont, Mark , month = sep, year =. Handbook of

  72. [75]

    Tony and Guo, Zijian and Ma, Rong , month = apr, year =

    Variable Selection for Global. Journal of the American Statistical Association , author =. 2023 , keywords =. doi:10.1080/01621459.2021.1969240 , number =

    work page doi:10.1080/01621459.2021.1969240 2023

  73. [76]

    Journal of Microbiology , author =

    Machine learning methods for microbiome studies , volume =. Journal of Microbiology , author =. 2020 , keywords =. doi:10.1007/s12275-020-0066-8 , language =

    work page doi:10.1007/s12275-020-0066-8 2020

  74. [77]

    PLOS Biology , author =

    Human microbiome research:. PLOS Biology , author =. 2023 , keywords =. doi:10.1371/journal.pbio.3002053 , language =

    work page doi:10.1371/journal.pbio.3002053 2023

  75. [78]

    Nature Communications , author =

    Microbiome differential abundance methods produce different results across 38 datasets , volume =. Nature Communications , author =. 2022 , keywords =. doi:10.1038/s41467-022-28034-z , language =

    work page doi:10.1038/s41467-022-28034-z 2022

  76. [79]

    Microbiome , author =

    Identifying biases and their potential solutions in human microbiome studies , volume =. Microbiome , author =. 2021 , keywords =. doi:10.1186/s40168-021-01059-0 , number =

    work page doi:10.1186/s40168-021-01059-0 2021

  77. [80]

    Nature Methods , author =

    Accessible, curated metagenomic data through. Nature Methods , author =. 2017 , pages =. doi:10.1038/nmeth.4468 , language =

    work page doi:10.1038/nmeth.4468 2017

  78. [81]

    Nature , author =

    A metagenome-wide association study of gut microbiota in type 2 diabetes , volume =. Nature , author =. 2012 , keywords =. doi:10.1038/nature11450 , language =

    work page doi:10.1038/nature11450 2012

  79. [82]

    Scandinavian Journal of Statistics , author =

    Bayesian. Scandinavian Journal of Statistics , author =. 2003 , keywords =. doi:10.1111/1467-9469.00349 , language =

    work page doi:10.1111/1467-9469.00349 2003

  80. [83]

    Journal of the Royal Statistical Society Series B , author =

    A Note on Least-Squares Approximation in the Bayesian Analysis of Regression Models , volume =. Journal of the Royal Statistical Society Series B , author =. 1984 , pages =. doi:10.1111/j.2517-6161.1984.tb01285.x , number =

    work page doi:10.1111/j.2517-6161.1984.tb01285.x 1984

Showing first 80 references.