Pith. sign in

REVIEW 4 minor 99 references

An R package that turns dynamic quantile linear models into an integrated Bayesian workflow: specify, fit with MCMC or fast variational Bayes, forecast, and diagnose from one object.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 06:33 UTC pith:OGWZCVCQ

load-bearing objection A solid, honest software paper—LDVB is the new piece, the package looks usable, but the benchmark is a single seed and LDVB interval accuracy is never checked against MCMC.

arxiv 2607.22760 v2 pith:OGWZCVCQ submitted 2026-07-23 stat.ME

exdqlm: An R Package for Estimation and Analysis of Flexible Dynamic Quantile Linear Models

Antonio De Leon , Raquel Barata , Raquel Prado , Bruno Sans\'o This is my paper
classification stat.ME MSC 62F1562M10
keywords dynamic quantile regressionextended asymmetric Laplacestate-space modelvariational BayesLaplace-delta approximationMCMCtransfer functionR package
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper presents exdqlm, an R package for Bayesian quantile regression focused on dynamic state-space models for time series. It claims to close a practical gap: existing quantile-regression software handles time only through observed predictors, while Gaussian state-space software lacks quantile likelihoods. The package builds on extended dynamic quantile linear models (exDQLMs), which use the extended asymmetric Laplace (exAL) error family, and provides both MCMC posterior simulation and a faster Laplace-delta variational Bayes (LDVB) approximation. A sympathetic reader would care because this makes time-varying quantile estimation, forecasting, diagnostics, transfer-function covariate effects, and synthesis across quantile levels available through one R interface.

Core claim

The core discovery is that the exAL-based exDQLM can be delivered as a practical, unified software workflow. The exAL distribution is defined so that the mass of the error below zero equals the target quantile p0, making the state contribution F_t^T θ_t the p0-quantile of the response. The exAL admits a hierarchical latent-variable representation that turns the observation equation into a Gaussian one conditional on latent variables; this property is what lets the package reuse standard Kalman recursions for the state sequence while handling the nonconjugate scale–skewness block with slice sampling or a Laplace-delta variational approximation. The paper demonstrates the workflow on Lake Huro

What carries the argument

The central object is the extended asymmetric Laplace (exAL) distribution, a parametric generalization of the asymmetric Laplace family with a skewness parameter γ restricted to an interval determined by p0. Its defining property, that the integral of the exAL density from −∞ to 0 equals p0, makes any linear combination F_t^T θ_t in the observation equation equal to the p0-quantile of the response. The exAL has a mixture representation over two latent variables—an exponential v_t and a positive truncated normal s_t—so conditional on these, the observation error is Gaussian. This turns posterior inference into a standard Gaussian state-space problem for the latent state, with the scale and sk

Load-bearing premise

The claim that the fitted state is the p0-quantile of the response rests entirely on the exAL distribution's property that the mass below zero equals p0; if that property or the mixture representation behind it fails for some parameter values, the model's quantile interpretation collapses.

What would settle it

Compute the integral of the exAL density from −∞ to 0 to high precision across a dense grid of p0 and γ within the admissible interval (L,U); a deviation from p0 anywhere would invalidate the quantile interpretation. Alternatively, simulate from the mixture representation and check that the empirical proportion of values ≤0 converges to p0; a mismatch would show the augmentation used by MCMC and LDVB is not faithful.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Users can obtain time-varying quantile estimates, credible intervals, and k-step-ahead forecasts from a single fitted object, choosing between full MCMC and the faster LDVB approximation.
  • Transfer-function specifications allow covariates to affect a quantile through a decaying lag structure, with a summary of the decay horizon reported as median.kt.
  • Separately fitted quantile models can be combined into one posterior predictive distribution via quantileSynthesis(), with optional monotonicity corrections for quantile crossing.
  • CRPS, PPLC, PIT, and KL diagnostics give a reproducible basis for comparing models and selecting discount factors.
  • Static AL/exAL regression with shrinkage priors is supported through the same interface, with LDVB running substantially faster than MCMC in the presented benchmark.

Where Pith is reading between the lines

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

  • A natural extension the paper leaves implicit is to let the transfer-function decay rate λ be estimated with a prior rather than selected from a fixed grid; the package's fixed-λ workflow could be amended with a joint Bayesian update.
  • The large LDVB speed advantage suggests it could be promoted from initialization to a default model-screening tool, with MCMC reserved for final tail-sensitive inference; the paper only uses LDVB as a warm start.
  • The exAL augmentation may carry over to multivariate or nonlinear dynamic quantile models, since the latent-mixture trick that linearizes the observation equation is not tied to the linear state-space form.
  • Diagnostic-based discount-factor selection could be formalized as continuous optimization over the discount-factor space, a modest extension of the coarse-grid scan demonstrated in the sunspot example.

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

0 major / 4 minor

Summary. The paper presents the R package exdqlm, which implements Bayesian quantile regression with a focus on dynamic state-space models. The package is built around the extended asymmetric Laplace (exAL) distribution and the extended dynamic quantile linear model (exDQLM) framework, providing MCMC and Laplace-delta variational Bayes (LDVB) inference, transfer-function state augmentation, static exAL/AL regression with shrinkage priors, forecasting, posterior-predictive synthesis, and diagnostics. The manuscript describes the model formulation, package design, and four worked examples, with appendices collecting posterior targets and implementation details. The central claim is that exdqlm fills a practical gap between general quantile-regression software and Gaussian state-space software by supplying a dedicated Bayesian dynamic quantile workflow.

Significance. If the claims are correct, the package is a useful and well-scoped software contribution. Its main strengths are transparency and reproducibility: the posterior targets in Appendices A-F are coherent, the examples exercise the claimed interfaces, the limitations of LDVB, one-quantile-at-a-time fitting, and post hoc synthesis are explicitly stated, and the replication scripts, seeds, and backend profiles are documented. The paper does not introduce new statistical theory; rather, it packages and integrates methods from Yan et al. (2025) and Barata et al. (2022) into a usable R workflow. This is appropriate for a Journal of Statistical Software submission, and the paper clearly delineates the package's niche relative to existing quantile-regression and state-space software.

minor comments (4)
  1. [Section 2.1, Eq. (3) / Appendix A.1] The central quantile interpretation rests on the imported exAL property that the CDF of the exAL density at the location parameter equals p0 over the admissible gamma interval. Since this property is load-bearing and the package exposes dexal(), pexal(), qexal(), and rexal(), the manuscript would be strengthened by a small numerical verification that the implemented density/CDF/quantile functions are mutually consistent and that the stated integral identity holds over a grid of (p0, sigma, gamma). This is a verification gap, not a demonstrated error.
  2. [Section 2.2, Eqs. (11)-(12)] The PIT, standardized forecast errors, and KL diagnostics are computed under a Gaussian MAP plug-in approximation to the one-step-ahead predictive distribution. Because the exAL/AL error distribution is non-Gaussian, these diagnostics can reflect the approximation error as well as model misfit. The text already calls them 'diagnostic summaries,' but I recommend an explicit sentence noting that the exact PIT would require integration over latent variables and parameters, and that the reported values are approximate calibration summaries.
  3. [Section 4.4, Table 10] The static simulation benchmark uses a single data realization, so the reported differences between LDVB and MCMC (e.g., active RMSE 0.264 vs 0.168 at p0=0.25) may be within Monte Carlo noise. Adding repeated-seed summaries with standard errors or boxplots would make the benchmark claims more robust and would better support the qualitative statement that LDVB is a useful first-pass screen.
  4. [Section 4.2, Table 7 and Section 4.4] The LDVB-vs-MCMC comparison focuses on point summaries, predictive scores, and runtimes. Since one of the package's advertised features is approximate posterior uncertainty quantification, it would be informative to include at least one quantitative comparison of LDVB and MCMC credible intervals, such as interval lengths or coverage in a repeated-setting simulation. This would help users calibrate their expectations when using LDVB for interval summaries.

Circularity Check

0 steps flagged

No circularity: software contribution built on cited external and prior work, with held-out and known-target validation.

full rationale

The paper's stated contribution is software-centered: it packages the exAL family of Yan et al. (2025) and the exDQLM framework of Barata et al. (2022) into an integrated R workflow. The exAL mixture representation (Eq. 3) and the quantile identity are imported from external published work, not rederived as a new result, so there is no self-definitional step in the paper itself. The exDQLM state-space formulation comes from prior work by two of the current authors, but it is presented as a foundation, not as an external uniqueness theorem, and the current paper's claims concern package capability rather than a new derivation. No fitted input is relabeled as a prediction: discount-factor selection using CRPS is explicitly described as in-sample diagnostic screening; the transfer-function rate lambda is selected on training-period PPLC and then evaluated on a held-out 18-month window; the static simulation benchmark compares estimates to a known sparse quantile signal on an independent holdout. Limitations such as LDVB approximation, one-quantile-at-a-time fitting, post hoc synthesis, and fixed lambda are disclosed in Section 5. The only nontrivial imported premise is the exAL CDF property from Yan et al. (2025); that is a verification gap external to this paper, not a circular reduction.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The paper's central claim is about software functionality, so the free parameters are the user-set or grid-selected tuning choices in the examples, not fitted constants that underpin a new statistical derivation. The main domain assumptions come from the exAL distribution and the variational approximation, both taken from prior literature.

free parameters (5)
  • Transfer rate lambda = 0.85
    Selected from grid {0.70, 0.75, 0.80, 0.85, 0.90, 0.95, 0.99} by minimizing training PPLC in Section 4.3.
  • Seasonal discount factor (sunspot example) = 0.85
    Selected from grid {0.85, 0.90, 0.95, 1.00} by minimizing in-sample CRPS in Section 4.2.
  • Trend discount factor (examples) = 0.9
    Hand-set in Lake Huron (0.9) and sunspot (0.9) examples; not fitted but affects the evolution covariance.
  • Prior hyperparameters (a_sig, b_sig, m_gam, s_gam, df_gam) = defaults (2.1, 1.1, 0, 1, 1)
    Default hyperparameters in Table 2 are chosen by hand and used in all examples unless overridden.
  • Initial state prior m0 = e.g., 579 for Lake Huron, 50 for sunspots
    Domain-scale values chosen before fitting, not estimated from the data.
axioms (4)
  • domain assumption The exAL error distribution satisfies the quantile property ∫_{-∞}^0 exAL_{p0}(ε|0,σ,γ)dε = p0 and admits the mixture representation in Eq. (3).
    Taken from Yan et al. (2025); the entire quantile interpretation and the computational augmentation depend on this property.
  • standard math Conditional on latent variables, the exDQLM becomes a Gaussian state-space model, so standard Kalman filtering, smoothing, and backward-sampling recursions are valid.
    Used throughout Appendix B for state updates; relies on well-known Gaussian state-space results.
  • domain assumption The Laplace-delta variational approximation of Wang and Blei (2013) yields accurate posterior moments for the nonconjugate (σ,γ) block.
    LDVB uses delta-method expectations (Appendix C); the paper does not provide a formal error bound or convergence guarantee.
  • ad hoc to paper The one-step-ahead predictive distribution used in diagnostics is adequately approximated by a Gaussian with MAP plug-in moments.
    Section 2.2 defines PIT and standardized errors using MAP plug-in location and variance; if the predictive is far from Gaussian, these diagnostics can be systematically misleading.

pith-pipeline@v1.3.0-alltime-deepseek · 33910 in / 9972 out tokens · 97929 ms · 2026-08-01T06:33:38.590580+00:00 · methodology

0 comments
read the original abstract

We present the R package exdqlm for Bayesian quantile regression, with primary emphasis on dynamic state-space quantile models for time series. The package is built around extended dynamic quantile linear models (exDQLMs), which use the extended asymmetric Laplace (exAL) family, a parametric extension of the asymmetric Laplace (AL) distribution commonly used in quantile regression. The software provides posterior simulation via Markov chain Monte Carlo (MCMC) and fast approximate posterior inference via Laplace-delta variational Bayes (LDVB), supporting posterior uncertainty quantification while also providing a computationally efficient option for longer time series. The same package interface supports static exAL quantile regression with regularized priors, dynamic transfer-function models for nonlinear input effects at a given quantile, post hoc posterior-predictive synthesis across separately fitted quantiles, forecasting, and quantitative and visual diagnostics for model evaluation.

Figures

Figures reproduced from arXiv: 2607.22760 by Antonio De Leon, Bruno Sans\'o, Raquel Barata, Raquel Prado.

Figure 1
Figure 1. Figure 1: Example 1: Lake Huron. Trace and density plots from the median run with 7000 [PITH_FULL_IMAGE:figures/full_fig_p016_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example 1: Lake Huron. Top-left: posterior mean estimates and 95% CrIs of the [PITH_FULL_IMAGE:figures/full_fig_p017_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example 2: Sunspots. Top: The sunspot time series from 1700 to 1988. Bottom [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example 2: Sunspots. The QQ plots (left column), ACF plots of the PIT sequence [PITH_FULL_IMAGE:figures/full_fig_p022_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example 3: Big Tree water flow. Top: observed average monthly water flow at the [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Example 3: Big Tree water flow. Top: posterior mean estimates and 95% CrIs of the [PITH_FULL_IMAGE:figures/full_fig_p027_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Example 3: Big Tree water flow. Posterior mean estimates and 95% CrIs of the [PITH_FULL_IMAGE:figures/full_fig_p028_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Example 3: Big Tree water flow. Held-out 18-month conditional forecast display [PITH_FULL_IMAGE:figures/full_fig_p033_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Example 4: sparse static exAL regression under the correlated Gaussian benchmark [PITH_FULL_IMAGE:figures/full_fig_p035_9.png] view at source ↗

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

99 extracted references · 3 canonical work pages

  1. [1]

    Fast inference for time-varying quantiles via flexible dynamic models with application to the characterization of atmospheric rivers

    Barata R, Prado R, Sans\' o B (2022). Fast inference for time-varying quantiles via flexible dynamic models with application to the characterization of atmospheric rivers . The Annals of Applied Statistics, 16(1), 247--271. doi:10.1214/21-AOAS1497

    work page doi:10.1214/21-aoas1497 2022

  2. [2]

    Statistical Inference Under Order Restrictions: The Theory and Application of Isotonic Regression

    Barlow RE, Bartholomew DJ, Bremner JM, Brunk HD (1972). Statistical Inference Under Order Restrictions: The Theory and Application of Isotonic Regression. John Wiley & Sons, New York

    1972

  3. [3]

    bayesQR : Bayesian Quantile Regression

    Benoit D, Al-Hamzawi R, Yu K, Van den Poel D (2023). bayesQR : Bayesian Quantile Regression . R package version 2.4, ://CRAN.R-project.org/package=bayesQR

    2023

  4. [4]

    brms : Special Family Functions for brms Models

    B \"u rkner PC (2026). brms : Special Family Functions for brms Models . Package documentation, accessed 2026-04-19, ://paulbuerkner.com/brms/reference/brmsfamily.html

    2026

  5. [5]

    On Gibbs sampling for state space models

    Carter CK, Kohn R (1994). On Gibbs sampling for state space models. Biometrika, 81(3), 541--553

    1994

  6. [6]

    Quantile and Probability Curves Without Crossing

    Chernozhukov V, Fern \'a ndez-Val I, Galichon A (2010). Quantile and Probability Curves Without Crossing. Econometrica, 78(3), 1093--1125. doi:10.3982/ECTA7880

    doi:10.3982/ecta7880 2010

  7. [7]

    pqrBayes : Bayesian Penalized Quantile Regression

    Fan K, Wu C, Ren J, Li X, Zhou F (2026). pqrBayes : Bayesian Penalized Quantile Regression . R package version 1.2.1, ://CRAN.R-project.org/package=pqrBayes

    2026

  8. [8]

    qgam : Smooth Additive Quantile Regression Models

    Fasiolo M, Griffiths B (2025). qgam : Smooth Additive Quantile Regression Models . R package version 2.0.0, ://CRAN.R-project.org/package=qgam

    2025

  9. [9]

    Data augmentation and dynamic linear models

    Fr \"u hwirth-Schnatter S (1994). Data augmentation and dynamic linear models. Journal of time series analysis, 15(2), 183--202

    1994

  10. [10]

    qrcm : Quantile Regression Coefficients Modeling

    Frumento P (2025). qrcm : Quantile Regression Coefficients Modeling . R package version 3.2, ://CRAN.R-project.org/package=qrcm

    2025

  11. [11]

    lqr : Robust Linear Quantile Regression

    Galarza CE, Benites L, Bourguignon M, Lachos VH (2024). lqr : Robust Linear Quantile Regression . R package version 5.2, ://CRAN.R-project.org/package=lqr

    2024

  12. [12]

    Model choice: a minimum posterior predictive loss approach

    Gelfand AE, Ghosh SK (1998). Model choice: a minimum posterior predictive loss approach. Biometrika, 85(1), 1--11

    1998

  13. [13]

    Strictly Proper Scoring Rules, Prediction, and Estimation

    Gneiting T, Raftery AE (2007). Strictly Proper Scoring Rules, Prediction, and Estimation. Journal of the American Statistical Association, 102(477), 359--378. doi:10.1198/016214506000001437

    doi:10.1198/016214506000001437 2007

  14. [15]

    conquer : Convolution-Type Smoothed Quantile Regression

    He X, Pan X, Tan KM, Zhou WX (2023). conquer : Convolution-Type Smoothed Quantile Regression . R package version 1.3.3, ://CRAN.R-project.org/package=conquer

    2023

  15. [16]

    Time series modeling via hierarchical mixtures

    Huerta G, Jiang W, Tanner MA (2003). Time series modeling via hierarchical mixtures. Statistica Sinica, pp. 1097--1118

    2003

  16. [17]

    Quantile regression

    Koenker R (2005). Quantile regression. Cambridge University Press, New York

    2005

  17. [18]

    quantreg : Quantile Regression

    Koenker R (2025). quantreg : Quantile Regression . R package version 6.1, ://CRAN.R-project.org/package=quantreg

    2025

  18. [19]

    Bayesian semiparametric modelling in quantile regression

    Kottas A, Krnjaji \'c M (2009). Bayesian semiparametric modelling in quantile regression. Scandinavian Journal of Statistics, 36(2), 297--319

    2009

  19. [20]

    Gibbs sampling methods for B ayesian quantile regression

    Kozumi H, Kobayashi G (2011). Gibbs sampling methods for B ayesian quantile regression. Journal of statistical computation and simulation, 81(11), 1565--1578

    2011

  20. [21]

    On information and sufficiency

    Kullback S, Leibler RA (1951). On information and sufficiency. The annals of mathematical statistics, 22(1), 79--86

    1951

  21. [22]

    Verification Tools for Probabilistic Forecasts of Continuous Hydrological Variables

    Laio F, Tamea S (2007). Verification Tools for Probabilistic Forecasts of Continuous Hydrological Variables. Hydrology and Earth System Sciences, 11(4), 1267--1277. doi:10.5194/hess-11-1267-2007

    doi:10.5194/hess-11-1267-2007 2007

  22. [23]

    quantregGrowth : Non-Crossing Additive Regression Quantiles and Non-Parametric Growth Charts

    Muggeo VMR (2025). quantregGrowth : Non-Crossing Additive Regression Quantiles and Non-Parametric Growth Charts . R package version 1.7-2, ://CRAN.R-project.org/package=quantregGrowth

    2025

  23. [24]

    Slice sampling

    Neal RM (2003). Slice sampling. The Annals of Statistics, 31(3), 705--767. doi:10.1214/aos/1056562461

    arXiv 2003

  24. [25]

    Shrinkage with shrunken shoulders: Gibbs sampling shrinkage model posteriors with guaranteed convergence rates

    Nishimura A, Suchard MA (2023). Shrinkage with shrunken shoulders: Gibbs sampling shrinkage model posteriors with guaranteed convergence rates. Bayesian Analysis, 18(2), 367--390. doi:10.1214/22-BA1308

    doi:10.1214/22-ba1308 2023

  25. [26]

    Climate Indices: Monthly Atmospheric and Ocean Time Series

    NOAA Physical Sciences Laboratory (2026). Climate Indices: Monthly Atmospheric and Ocean Time Series. ://psl.noaa.gov/data/climateindices/

    2026

  26. [27]

    dynr : Dynamic Models with Regime-Switching

    Ou L, Hunter M, Chow SM, Ji L, Chen M, Hung HJ, Lee J, Li Y, Park J (2021). dynr : Dynamic Models with Regime-Switching . R package version 0.1.16-2, ://CRAN.R-project.org/package=dynr

    2021

  27. [28]

    dlm : Bayesian and Likelihood Analysis of Dynamic Linear Models

    Petris G, Gilks W (2018). dlm : Bayesian and Likelihood Analysis of Dynamic Linear Models . R package version 1.1-5, ://CRAN.R-project.org/package=dlm

    2018

  28. [29]

    CODA: Convergence Diagnosis and Output Analysis for MCMC

    Plummer M, Best N, Cowles K, Vines K (2006). CODA: Convergence Diagnosis and Output Analysis for MCMC. R News, 6(1), 7--11. ://journal.r-project.org/archive/

    2006

  29. [30]

    Time Series: Modeling, Computation and Inference

    Prado R, Ferreira MA, West M (2021). Time Series: Modeling, Computation and Inference. 2nd edition. Chapman and Hall/CRC Press

    2021

  30. [31]

    Multivariate time series modeling and classification via hierarchical VAR mixtures

    Prado R, Molina F, Huerta G (2006). Multivariate time series modeling and classification via hierarchical VAR mixtures. Computational Statistics & Data Analysis, 51(3), 1445--1462

    2006

  31. [32]

    R: A Language and Environment for Statistical Computing

    R Core Team (2026). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. ://www.R-project.org/

    2026

  32. [33]

    Flexible B ayesian quantile regression for independent and clustered data

    Reich BJ, Bondell HD, Wang HJ (2009). Flexible B ayesian quantile regression for independent and clustered data. Biostatistics, 11(2), 337--352

    2009

  33. [34]

    Remarks on a multivariate transformation

    Rosenblatt M (1952). Remarks on a multivariate transformation. The annals of mathematical statistics, 23(3), 470--472

    1952

  34. [35]

    rqPen : Penalized Quantile Regression

    Sherwood B, Li S, Maidman A (2026). rqPen : Penalized Quantile Regression . R package version 4.2, ://CRAN.R-project.org/package=rqPen

    2026

  35. [36]

    A B ayesian nonparametric approach to inference for quantile regression

    Taddy MA, Kottas A (2010). A B ayesian nonparametric approach to inference for quantile regression. Journal of Business & Economic Statistics, 28(3), 357--369

    2010

  36. [37]

    qrjoint : Joint Estimation in Linear Quantile Regression

    Tokdar S, Cunningham E (2025). qrjoint : Joint Estimation in Linear Quantile Regression . R package version 2.0-11, ://CRAN.R-project.org/package=qrjoint

    2025

  37. [38]

    National Water Information System data available on the World Wide Web (USGS Water Data for the Nation)

    US Geological Survey (2016). National Water Information System data available on the World Wide Web (USGS Water Data for the Nation). ://waterdata.usgs.gov/nwis/

    2016

  38. [39]

    Modern Applied Statistics with S

    Venables WN, Ripley BD (2002). Modern Applied Statistics with S . Fourth edition. Springer, New York. ISBN 0-387-95457-0, ://www.stats.ox.ac.uk/pub/MASS4/

    2002

  39. [40]

    Variational Inference in Nonconjugate Models

    Wang C, Blei DM (2013). Variational Inference in Nonconjugate Models. Journal of Machine Learning Research, 14, 1005--1031. ://jmlr.org/papers/v14/wang13b.html

    2013

  40. [41]

    Bayesian Forecasting and Dynamic Models

    West M, Harrison J (1997). Bayesian Forecasting and Dynamic Models. 2 edition. Springer, New York. doi:10.1007/b98971

    work page doi:10.1007/b98971 1997

  41. [42]

    A New Family of Error Distributions for Bayesian Quantile Regression

    Yan Y, Zheng X, Kottas A (2025). A New Family of Error Distributions for Bayesian Quantile Regression. Bayesian Analysis. doi:10.1214/25-BA1507

    work page doi:10.1214/25-ba1507 2025

  42. [43]

    Bayesian quantile regression

    Yu K, Moyeed RA (2001). Bayesian quantile regression. Statistics & Probability Letters, 54(4), 437--447

    2001

  43. [44]

    2026 , url =

    R: A Language and Environment for Statistical Computing , author =. 2026 , url =

    2026

  44. [45]

    R News , year =

    CODA: Convergence Diagnosis and Output Analysis for MCMC , author =. R News , year =

  45. [46]

    W. N. Venables and B. D. Ripley , publisher =. Modern Applied Statistics with. 2002 , note =

    2002

  46. [47]

    Taddy, Matthew A and Kottas, Athanasios , journal=. A. 2010 , publisher=

    2010

  47. [48]

    , title =

    Gneiting, Tilmann and Raftery, Adrian E. , title =. Journal of the American Statistical Association , volume =. 2007 , doi =

    2007

  48. [49]

    Prado and M

    R. Prado and M. A. Ferreira and M. West , year=. Time Series: Modeling, Computation and Inference , publisher=

  49. [50]

    Hydrology and Earth System Sciences , volume =

    Laio, Francesco and Tamea, Stefania , title =. Hydrology and Earth System Sciences , volume =. 2007 , doi =

    2007

  50. [51]

    Journal of Mathematical Psychology , volume=

    A tutorial on variational Bayes for latent linear stochastic time-series models , author=. Journal of Mathematical Psychology , volume=. 2014 , publisher=

    2014

  51. [52]

    2003 , school=

    Variational algorithms for approximate Bayesian inference , author=. 2003 , school=

    2003

  52. [53]

    Scandinavian Journal of Statistics , volume=

    Bayesian semiparametric modelling in quantile regression , author=. Scandinavian Journal of Statistics , volume=. 2009 , publisher=

    2009

  53. [54]

    Flexible

    Reich, Brian J and Bondell, Howard D and Wang, Huixia J , journal=. Flexible. 2009 , publisher=

    2009

  54. [55]

    Benoit, Dries and Al-Hamzawi, Rahim and Yu, Keming and Van den Poel, Dirk , year =

  55. [56]

    Fan, Kun and Wu, Cen and Ren, Jie and Li, Xiaoxi and Zhou, Fei , year =

  56. [57]

    Statistical Inference Under Order Restrictions: The Theory and Application of Isotonic Regression , author =

  57. [58]

    Quantile and Probability Curves Without Crossing , journal =

    Chernozhukov, Victor and Fern. Quantile and Probability Curves Without Crossing , journal =. 2010 , doi =

    2010

  58. [59]

    Biometrika , volume=

    On Gibbs sampling for state space models , author=. Biometrika , volume=. 1994 , publisher=

    1994

  59. [60]

    Journal of time series analysis , volume=

    Data augmentation and dynamic linear models , author=. Journal of time series analysis , volume=. 1994 , publisher=

    1994

  60. [61]

    Solar Minimum; Solar Maximum , year =

    Karen Fox , journal =. Solar Minimum; Solar Maximum , year =

  61. [62]

    2016 , url =

    National Water Information System data available on the World Wide Web (USGS Water Data for the Nation) , title =. 2016 , url =

    2016

  62. [63]

    Journal of Geophysical Research: Atmospheres , volume=

    Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century , author=. Journal of Geophysical Research: Atmospheres , volume=. 2003 , publisher=

    2003

  63. [64]

    Climate Indices: Monthly Atmospheric and Ocean Time Series , year =

  64. [65]

    Journal of Statistical Software , volume=

    bayesQR: A Bayesian approach to quantile regression , author=. Journal of Statistical Software , volume=

  65. [66]

    Giovanni Petris and Wally Gilks , year =

  66. [67]

    Lu Ou and Michael Hunter and Sy-Min Chow and Linying Ji and Meng Chen and Hui-Ju Hung and Jungmin Lee and Yanling Li and Jonathan Park , year =

  67. [68]

    Striaukas, Jonas , year =

  68. [69]

    He, Xuming and Pan, Xiaoou and Tan, Kean Ming and Zhou, Wen-Xin , year =

  69. [70]

    Fasiolo, Matteo and Griffiths, Ben , year =

  70. [71]

    and Benites, Luis and Bourguignon, Marcelo and Lachos, Victor H

    Galarza, Christian E. and Benites, Luis and Bourguignon, Marcelo and Lachos, Victor H. , year =

  71. [72]

    Journal of Statistical Software , volume=

    An R package for dynamic linear models , author=. Journal of Statistical Software , volume=

  72. [73]

    Koenker, Roger , year =

  73. [74]

    Muggeo, Vito M. R. , year =

  74. [75]

    Frumento, Paolo , year =

  75. [76]

    Tokdar, Surya and Cunningham, Erika , year =

  76. [77]

    , year =

    Xu, Steven and Reich, Brian J. , year =

  77. [78]

    R package version 5.05 , author=

    quantreg: Quantile Regression. R package version 5.05 , author=. R Foundation for Statistical Computing: Vienna) Available at: http://CRAN. R-project. org/package= quantreg , year=

  78. [79]

    Sherwood, Ben and Li, Shaobo and Maidman, Adam , year =

  79. [80]

    Fogarty, Ben , year =

  80. [81]

    2026 , note =

    Working with Quantile Regression Models , author =. 2026 , note =

    2026

Showing first 80 references.