pith. sign in

arxiv: 2606.12164 · v1 · pith:RMPWMZCGnew · submitted 2026-06-10 · 📊 stat.ME

Bayesian Effect Selection for Additive Quantile Regression with an Application to Air Pollution Thresholds

Nadja Klein , Aaron Wei Qi Lee , Jorge Mateu This is my paper

Pith reviewed 2026-06-27 08:41 UTC · model grok-4.3

classification 📊 stat.ME
keywords Bayesian effect selectionadditive quantile regressionDemmler-Reinsch basisspike and slab priorsair pollution thresholdsNO2 concentrationsconsistent estimationlinear nonlinear decomposition
0
0 comments X

The pith

Demmler-Reinsch basis expansion decomposes additive effects into consistently estimable linear and nonlinear parts for Bayesian selection in quantile regression.

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

The paper develops Bayesian effect selection for additive quantile regression by replacing standard mixed-model spline representations with a Demmler-Reinsch basis that orthogonally separates each covariate effect into a linear component and a nonlinear component. Separate spike-and-slab priors are placed on scalar importance parameters for these two components, and an efficient Gibbs sampler performs the selection. Theoretical results establish consistent estimation of both components under the asymmetric Laplace working likelihood. Simulations confirm robustness to the induced misspecification and better performance than mixed-model approaches. The method is used on Madrid air-pollution data to show that extreme NO2 concentrations respond differently to climatological variables and traffic-related spatial structure.

Core claim

By employing a Demmler-Reinsch basis expansion in additive quantile regression, each additive effect can be orthogonally decomposed into linear and nonlinear parts that are both estimated consistently; Bayesian selection proceeds via separate spike-and-slab priors on the associated scalar importance parameters, implemented through an efficient Gibbs sampler, and this framework distinguishes the drivers of threshold exceedances in the Madrid NO2 data.

What carries the argument

Demmler-Reinsch basis expansion that orthogonally decomposes each additive effect into linear and nonlinear components, together with separate spike-and-slab priors on the scalar importance parameters of those components.

If this is right

  • Both the linear and nonlinear parts of each additive effect are estimated consistently.
  • Separate spike-and-slab priors allow data-driven decisions on whether an effect is absent, linear only, nonlinear only, or both.
  • The Gibbs sampler yields an efficient posterior for model selection under the quantile regression working likelihood.
  • In the air-pollution application the threshold-relevant NO2 levels are shown to be driven differently by climatological variables and traffic-related spatial structure.

Where Pith is reading between the lines

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

  • The orthogonal decomposition may improve interpretability of quantile effects in other environmental or health applications where distinguishing linear trends from nonlinear thresholds matters.
  • Because the selection operates on scalar parameters, the approach could be combined with variable-selection methods that act on groups of covariates.
  • The robustness to asymmetric Laplace misspecification suggests the procedure might remain useful even if a different working likelihood, such as a direct quantile loss, is substituted.
  • Extension to spatio-temporal quantile models would allow the same linear-versus-nonlinear selection for air-pollution fields evolving over time.

Load-bearing premise

The asymmetric Laplace working likelihood induces misspecification, yet the consistency and selection results are claimed to hold anyway.

What would settle it

Repeated simulations with known true linear and nonlinear effects in which the posterior inclusion probabilities fail to recover the correct components or the point estimates diverge from the true values would falsify the consistency claim.

Figures

Figures reproduced from arXiv: 2606.12164 by Aaron Wei Qi Lee, Jorge Mateu, Nadja Klein.

Figure 1
Figure 1. Figure 1: Results from univariate preliminary analyses with [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Results from univariate preliminary analyses with [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Boxplots of the out-of-sample quantile loss from 10-fold cross-validation for each [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Estimated posterior effects for fτ,j,nonlin (nonlinear parts) for τ P t0.6, 0.8, 0.9u (across columns 1–3) and for all nine covariates (row-wise). Shown are the posterior mean (solid lines) and 95% pointwise credible intervals (shaded area). 24 [PITH_FULL_IMAGE:figures/full_fig_p025_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Estimated posterior effects for fτ,j “ fτ,j,nonlin ` fτ,j,lin (linear parts+nonlinear parts) for τ P t0.6, 0.8, 0.9u (across columns 1–3) and for all nine covariates (row-wise). Shown are the posterior mean (solid lines) and 95% pointwise credible intervals (shaded area). 25 [PITH_FULL_IMAGE:figures/full_fig_p026_5.png] view at source ↗
read the original abstract

Air pollution regulatory limits are typically defined in terms of exceedances of concentration thresholds which are naturally related to conditional quantiles of the pollutant distribution and are therefore of direct relevance for assessing severe pollution events. At the same time, it is important to determine not only whether a covariate affects air pollution but also whether this effect is linear, nonlinear, or both. We address these issues by developing a Bayesian effect selection approach for additive quantile regression. While commonly used mixed model representations (MMRs) of penalized splines allow for flexible nonlinear effects, they do not provide a meaningful separation of linear and nonlinear effect components. We therefore employ a Demmler-Reinsch basis expansion, which yields an orthogonal decomposition of each additive effect into linear and nonlinear parts and show theoretically that both effect components can be estimated consistently. To facilitate data-driven model building, we propose Bayesian effect selection with separate spike and slab priors on the scalar importance parameters associated with the linear and nonlinear components and implement an efficient Gibbs sampler. Through simulation studies, we demonstrate robustness to the misspecification induced by the employed asymmetric Laplace working likelihood and show superior performance relative to the MMR. In a detailed analysis of air pollution data in Madrid, Spain we highlight the added value of flexibly modeling extreme nitrogen dioxide (NO$_2$) concentrations and reveal that threshold-relevant pollution levels are driven differently by climatological variables and traffic-related spatial structure. These findings underline the need for advanced statistical models that support short-term decision-making and help local authorities mitigate, or potentially prevent, exceedances of NO$_2$ concentration limits.

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

1 major / 2 minor

Summary. The paper develops a Bayesian effect selection procedure for additive quantile regression. It replaces standard mixed-model spline representations with a Demmler-Reinsch basis that orthogonally decomposes each additive effect into linear and nonlinear components, establishes theoretical consistency for both components, places separate spike-and-slab priors on the associated scalar importance parameters, and implements the resulting model via Gibbs sampling. Simulations are used to illustrate robustness to asymmetric Laplace likelihood misspecification and superiority over mixed-model representations; the method is then applied to threshold-relevant NO2 concentrations in Madrid.

Significance. If the consistency result survives the working-likelihood misspecification, the orthogonal decomposition together with component-wise selection supplies a practically useful advance for quantile regression applications that require explicit separation of linear and nonlinear effects. The Demmler-Reinsch construction and the separate spike-and-slab mechanism are clear methodological strengths that directly address a limitation of existing penalized-spline quantile models.

major comments (1)
  1. [Abstract and theoretical consistency section] Abstract and the section presenting the theoretical consistency result: the claim that both linear and nonlinear Demmler-Reinsch components are estimated consistently is central to the paper. All estimation proceeds under the asymmetric Laplace working likelihood, which is misspecified unless the conditional errors are exactly ALD. The abstract states that simulations demonstrate robustness, yet gives no indication that the consistency proof itself is derived under misspecification or under conditions that survive it. Clarification is required on whether the proof assumes correct specification of the working model.
minor comments (2)
  1. [theoretical results] The manuscript would benefit from an explicit statement of the precise conditions (e.g., on the error distribution or on the design) under which the consistency theorem is proved.
  2. [simulation studies] Simulation section: the data-generating processes used to demonstrate robustness should be described in sufficient detail (including the exact error distributions and sample sizes) to allow readers to assess how far the misspecification departs from the ALD.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading of the manuscript and for identifying this important point regarding the theoretical consistency result. We address the comment below.

read point-by-point responses

  1. Referee: [Abstract and theoretical consistency section] Abstract and the section presenting the theoretical consistency result: the claim that both linear and nonlinear Demmler-Reinsch components are estimated consistently is central to the paper. All estimation proceeds under the asymmetric Laplace working likelihood, which is misspecified unless the conditional errors are exactly ALD. The abstract states that simulations demonstrate robustness, yet gives no indication that the consistency proof itself is derived under misspecification or under conditions that survive it. Clarification is required on whether the proof assumes correct specification of the working model.

    Authors: We agree that clarification is needed. The consistency result for the linear and nonlinear Demmler-Reinsch components is derived under the assumption that the data are generated from the asymmetric Laplace distribution (correct specification of the working likelihood). The abstract and theoretical section currently do not make this assumption explicit. In the revised manuscript we will add a clear statement to this effect in both the abstract and the consistency section, while retaining the existing simulation results that demonstrate robustness under misspecification. revision: yes

Circularity Check

0 steps flagged

No circularity; derivation relies on external basis and independent theoretical/simulation claims

full rationale

The paper's core steps—adopting the Demmler-Reinsch orthogonal decomposition for linear/nonlinear separation, stating a consistency result for the components, and placing separate spike-and-slab priors—are presented as building on established spline techniques with a new Bayesian selection layer. The abstract explicitly separates the theoretical consistency claim from the simulation-based robustness check for the asymmetric Laplace likelihood. No quoted equation or self-citation reduces a claimed prediction or uniqueness result to a fitted input or prior self-work by construction. The method is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on standard statistical assumptions for quantile regression and basis expansions; no new entities postulated.

axioms (2)
  • domain assumption Asymmetric Laplace distribution serves as a valid working likelihood for quantile estimation despite misspecification.
    Explicitly invoked in abstract for the Gibbs sampler and robustness claims.
  • standard math Demmler-Reinsch basis provides orthogonal decomposition allowing consistent separate estimation of linear and nonlinear components.
    Core to the effect selection procedure and theoretical consistency result.

pith-pipeline@v0.9.1-grok · 5816 in / 1308 out tokens · 16434 ms · 2026-06-27T08:41:14.769292+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

86 extracted references · 7 canonical work pages

  1. [1]

    1989 , author =

    Generalized Linear Models , publisher =. 1989 , author =

    1989

  2. [2]

    and He, X

    Yang, Y. and He, X. , title =. Annals of Statistics , volume =

  3. [3]

    Bissiri, P. G. and Holmes, C. C. and Walker, S. G. , title =. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume =. 2016 , doi =

    2016

  4. [4]

    2025 , journal=

    Generative Regression with IQ-BART , author=. 2025 , journal=

    2025

  5. [5]

    Eilers, P. H. C. , year =. Comments on ``. Journal of the Royal Statistical Society: Series C (Applied Statistics) , file =

  6. [6]

    2024 , volume=

    Distributional Regression for Data Analysis , author=. 2024 , volume=. doi:10.1146/annurev-statistics-040722-053607 , journal=

    work page doi:10.1146/annurev-statistics-040722-053607 2024

  7. [7]

    Horowitz and Sokbae Lee , title =

    Joel L. Horowitz and Sokbae Lee , title =. Journal of the American Statistical Association , volume =. 2005 , doi =

    2005

  8. [8]

    2024 , journal=

    Bayesian effect selection in additive models with an application to time-to-event data , author=. 2024 , journal=

    2024

  9. [9]

    Stone , title =

    Charles J. Stone , title =. Annals of Statistics , volume =. 1986 , doi =

    1986

  10. [10]

    2022 , howpublished =

    EEA , title =. 2022 , howpublished =

    2022

  11. [11]

    2018 , howpublished =

    Hyndman, Rob and Athanasopoulos, George , title =. 2018 , howpublished =

    2018

  12. [12]

    Spatial Statistics , month = oct, year =

    Nag, Pratik and Sun, Ying and Reich, Brian , title =. Spatial Statistics , volume =. doi:https://doi.org/10.1016/j.spasta.2023.100773 , url =

    work page doi:10.1016/j.spasta.2023.100773 2023

  13. [13]

    Air Quality, Atmosphere and Health , volume =

    Pak, Unjin and Kim, Chungsong and Ryu, Unsok and Sok, Kyongjin and Pak, Sungnam , title =. Air Quality, Atmosphere and Health , volume =. doi:, url =

  14. [14]

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

    Panja, Madhurima and Chakraborty, Tanujit and Biswas, Anubhab and Deb, Soudeep , title =. Journal of the Royal Statistical Society Series A: Statistics in Society , volume =. doi:https://doi.org/10.1093/jrsssa/qnag010 , url =

    work page doi:10.1093/jrsssa/qnag010

  15. [15]

    International Statistical Review , volume =

    Yang, Yunwen and Wang, Huixia Judy and He, Xuming , title =. International Statistical Review , volume =. doi:https://doi.org/10.1111/insr.12114 , url =

    work page doi:10.1111/insr.12114

  16. [16]

    Journal of the American Statistical Association , volume =

    Fasiolo, Matteo and Wood, Simon N. and Zaffran, Margaux and Nedellec, Rapha. Fast calibrated additive quantile regression , volume =. 2021 , publisher =. doi:10.1080/01621459.2020.1725521 , URL =

    work page doi:10.1080/01621459.2020.1725521 2021

  17. [17]

    , year =

    Meinshausen, N. , year =. Quantile regression forests , pages =

  18. [18]

    Model selection in quantile regression models , pages =

    Alhamzawi, Rahim , year =. Model selection in quantile regression models , pages =. Journal of Applied Statistics , doi =

  19. [19]

    Variable selection in quantile regression via Gibbs sampling , pages =

    Alhamzawi, Rahim and Yu, Keming , year =. Variable selection in quantile regression via Gibbs sampling , pages =. Journal of Applied Statistics , doi =

  20. [20]

    Conjugate priors and variable selection for Bayesian quantile regression , pages =

    Alhamzawi, Rahim and Yu, Keming , year =. Conjugate priors and variable selection for Bayesian quantile regression , pages =. Computational Statistics

  21. [21]

    Bayesian variable selection in quantile regression , pages =

    Yu, Keming and Chen, Cathy and Reed, Craig and Dunson, David , year =. Bayesian variable selection in quantile regression , pages =

  22. [22]

    Variable selection in quantile regression , pages =

    Wu, Yichao and Liu, Yufeng , year =. Variable selection in quantile regression , pages =

  23. [23]

    and Kneib, T

    Fenske, N. and Kneib, T. and Hothorn, T. , year =. Identifying risk factors for severe childhood malnutrition by boosting additive quantile regression , pages =

  24. [24]

    Airquality-News , year =

  25. [25]

    Urban-NO2-Atlas , year =

  26. [26]

    and Brezger, A

    Belitz, C. and Brezger, A. and Klein, N. and Kneib, T. and Lang, S. and Umlauf, N. , title =. 2015 , note=

    2015

  27. [27]

    1990 , author =

    Generalized Additive Models , publisher =. 1990 , author =

    1990

  28. [28]

    and Beauchamp, J

    Mitchell, T.J. and Beauchamp, J. J. , title =. Journal of the American Statistical Association , year =

  29. [29]

    Statistical

    Clyde, Merlise and George, Edward I , title =. Statistical. 2004 , volume =

    2004

  30. [30]

    and McCulloch, R.E

    George, E.I. and McCulloch, R.E. , title =. Statistica Sinica , year =

  31. [31]

    and Sillanp\"a\"a, M

    O'Hara, R. and Sillanp\"a\"a, M. , title =. Bayesian Analysis , year =

  32. [32]

    and Reich, Brian J

    Bondell, Howard D. and Reich, Brian J. and Wang, Huixia , year =. Noncrossing quantile regression curve estimation , pages =. Biometrika , doi =

  33. [33]

    Eilers, Paul H. C. and Marx, Brian D. , year =. Flexible smoothing with. Statistical Science , doi =

  34. [34]

    and Dortet-Bernadet, J.-L

    Rodrigues, T. and Dortet-Bernadet, J.-L. and Fan, Y. , year =. Simultaneous fitting of. Computational Statistics

  35. [35]

    and Eilers, Paul H

    Schnabel, Sabine K. and Eilers, Paul H. C. , year =. Simultaneous estimation of quantile curves using quantile sheets , pages =. AStA Advances in Statistical Analysis , doi =

  36. [36]

    Bayesian inference for additive mixed quantile regression models , journal =

    Yu Ryan Yue and H. Bayesian inference for additive mixed quantile regression models , journal =. 2011 , doi =

    2011

  37. [37]

    2017 , author =

    Generalized Additive Models : An Introduction with R , publisher =. 2017 , author =

    2017

  38. [38]

    2013 , author =

    Regression - Models, Methods and Applications , publisher =. 2013 , author =

    2013

  39. [39]

    , title=

    Tsionas, E.G. , title=. Journal of Statistical Computation and Simulation , year=

  40. [40]

    Bayesian semiparametric modelling in quantile regression , journal =

    Kottas, ATHANASIOS and Krnjaji\'. Bayesian semiparametric modelling in quantile regression , journal =. doi:https://doi.org/10.1111/j.1467-9469.2008.00626.x , abstract =

    work page doi:10.1111/j.1467-9469.2008.00626.x 2008

  41. [41]

    and Bondell, Howard D

    Reich, Brian J. and Bondell, Howard D. and Wang, Huixia J. , title = ". Biostatistics , volume =. 2009 , month =. doi:10.1093/biostatistics/kxp049 , eprint =

    work page doi:10.1093/biostatistics/kxp049 2009

  42. [42]

    Gelfand , title =

    Kristian Lum and Alan E. Gelfand , title =. Bayesian Analysis , number =. 2012 , doi =

    2012

  43. [43]

    Journal of Statistical Computation and Simulation , volume =

    Hideo Kozumi and Genya Kobayashi , title =. Journal of Statistical Computation and Simulation , volume =. 2011 , publisher =

    2011

  44. [44]

    Technical Report available at http://bura.brunel.ac.uk/handle/2438/3593 , year =

    Craig Reed and Keming Yu , title =. Technical Report available at http://bura.brunel.ac.uk/handle/2438/3593 , year =

  45. [45]

    Moyeed , keywords =

    Keming Yu and Rana A. Moyeed , keywords =. Bayesian quantile regression , journal =. 2001 , doi =

    2001

  46. [46]

    and Kneib, T

    Fahrmeir, L. and Kneib, T. and Lang, S. , title =. Statistica Sinica , year =

  47. [47]

    and Kneib, T

    Waldmann, E. and Kneib, T. and Yue, Y. R. and Lang, S. and Flexeder, C. , title =. Statistical Modelling , volume =. 2013 , doi =

    2013

  48. [48]

    and Kneib, T

    Klein, N. and Kneib, T. and Klasen, S. and Lang, S. , title =. 2015 , volume =

    2015

  49. [49]

    and Fahrmeir, L

    Scheipl, F. and Fahrmeir, L. and Kneib, T. , title =. 2012 , journal =

    2012

  50. [50]

    and Held, L

    Rue, H. and Held, L. , title =

  51. [51]

    2014 , journal =

    Multilevel structured additive regression , author =. 2014 , journal =

    2014

  52. [52]

    2018 , note =

    sdPrior: Scale-Dependent hyperpriors in structured additive distributional regression , author =. 2018 , note =

    2018

  53. [53]

    and Brezger, A

    Lang, S. and Brezger, A. , title =. Journal of Computational and Graphical Statistics , year =

  54. [54]

    and Kneib, T

    Klein, N. and Kneib, T. , title =. 2016 , volume =

    2016

  55. [55]

    and Rue, H

    Simpson, D. and Rue, H. and Riebler, A. and Martins, T. G. and S. Penalising model component complexity: A principled, practical approach to constructing priors , year =. Statistical Science , volume =

  56. [56]

    Additive

    David Rossell and Francisco Javier Rubio , year=. Additive

  57. [57]

    Bayesian Analysis , number =

    Nadja Klein and Manuel Carlan and Thomas Kneib and Stefan Lang and Helga Wagner , title =. Bayesian Analysis , number =. 2021 , doi =

    2021

  58. [58]

    2005 , author =

    Quantile Regression , publisher =. 2005 , author =

    2005

  59. [59]

    , title =

    Koenker, R. , title =. Advances in Social Science Research Using R , publisher =. 2010 , editor =

    2010

  60. [60]

    , title =

    Koenker, R. , title =. Brazilian Journal of Probability and Statistics , year =

  61. [61]

    and Bassett, G

    Koenker, R. and Bassett, G. , title =. Econometrica , year =

  62. [62]

    and Brauer, M

    Achakulwisut, P. and Brauer, M. and Hystad, P. and Anenberg, S.C. , year =. Global, national, and urban burdens of paediatric asthma incidence attributable to ambient. Lancet Planet Health , doi =

  63. [63]

    and Pocajt, V

    Antanasijevic, D. and Pocajt, V. and Peric-Grujic, A. and Ristic, M. , year =. Multiple-input-multiple-output general regression neural networks model for the simultaneous estimation of traffic-related air pollutant emissions , pages =. Atmospheric Pollution Research , doi =

  64. [64]

    WHO , year =. , doi =

  65. [65]

    and Bierlaire, M

    Bergantino, A.S. and Bierlaire, M. and Catalano, M. and Migliore, M. and Amoroso, S. , year =. Taste heterogeneity and latent preferences in the choice behaviour of freight transport operators , pages =. Transport Policy , doi =

  66. [66]

    and Santiago, J.L

    Borge, R. and Santiago, J.L. and de la Paz, D. and Martin, F. and Domingo, J. and Valdes, C. and Sanchez, B. and Rivas, E. and Rozas, M.T. and Lazaro, S. and Perez, J. and Fernandez, A. , year =. Application of a short term air quality action plan in. Science of Total Environment , doi =

  67. [67]

    and Galatioto, F

    Catalano, M. and Galatioto, F. and Bell, M. and Namdeo, A. and Bergantino, A.S. , year =. Improving the prediction of air pollution peak episodes generated by urban transport networks , pages =. Environmental Science and Policy , doi =

  68. [68]

    Air quality in

    EEA , year =. Air quality in. European Environment Agency Technical report no 12/2018 , doi =

    2018

  69. [69]

    and Valois, M.F

    Hatzopoulou, M. and Valois, M.F. and Levy, I. and Mihele, C. and Lu, G. and Bagg, S. and Minet, L. and Brook, J. , year =. Robustness of land-use regression models developed from mobile air pollutant measurements , pages =. Environmental Science Technology , doi =

  70. [70]

    and An, S.S

    Lee, D.H. and An, S.S. and Song, H.M. and Park, O.H. and Park, K.S. and Seo, G.Y. and Cho, Y.G. and Kim, E.S. , year =. The effect of traffic volume on the air quality at monitoring sites in. Korean Society of Environmental Health , doi =

  71. [71]

    and Schmitz, O

    Lu, M. and Schmitz, O. and de Hoogh, K. and Kai, Q. and Karssenberg, D. , year =. Evaluation of different methods and data sources to optimise modelling of. Environment International , doi =

  72. [72]

    and Chen, L

    Meng, X. and Chen, L. and Cai, J. and Zou, B. and Wu, C.-F. and Fu, Q. and Zhang, Y. and Liu, Y. and Kan, H. , year =. A land use regression model for estimating the. Environmental Research , doi =

  73. [73]

    and Nucifora, A.F.M

    Nunnari, G. and Nucifora, A.F.M. and Randieri, C. , year =. The application of neural techniques to the modelling of time-series of atmospheric pollution data , pages =. Ecological Modelling , doi =

  74. [74]

    and Trier, A

    Perez, P. and Trier, A. , year =. Prediction of. Atmospheric Environment , doi =

  75. [75]

    and Yi, J.S

    Prybutok, V.R. and Yi, J.S. and Mitchell, D. , year =. Comparison of neural network models with ARIMA and regression models for prediction of. European Journal of Operations Research , doi =

  76. [76]

    and Park, C

    Ryu, J. and Park, C. and Jeon, S.W. , year =. Mapping and statistical analysis of. Sustainability , doi =

  77. [77]

    and Pinardi, A

    Valks, P. and Pinardi, A. and Richter, A. and Lambert, J.-C. and Hao, N. and Loyola, D. and van Roozendael, M. and Emmadi, S. , year =. Operational total and tropospheric. Atmospheric Measurement Techniques , doi =

  78. [78]

    and Zhu, J

    Li, Y. and Zhu, J. , title =. Journal of Computational and Graphical Statistics , year =

  79. [79]

    and Zhang, H

    Lin, Y. and Zhang, H. , title =. Annals of Statistics , year =

  80. [80]

    and Bondell, H

    Lin, C.-Y. and Bondell, H. and Zhang, H. H. and Zou, H. , title =. Stat , year =

Showing first 80 references.