Pith. sign in

REVIEW 3 major objections 4 minor 37 references

Modeling Joint Health Effects of Environmental Exposure Mixtures with Bayesian Additive Regression Trees

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Soft BART inside a negative binomial model can recover nonlinear, interacting exposure-mixture surfaces for count outcomes, and ALE summaries of the Atlanta asthma data show ozone associated with more emergency-department visits and NO2…

desk verdict Solid applied demonstration of soft BART + ALE for count mixtures, but the headline findings rest on untested confounding assumptions; the simulation alone cannot carry them. read the letter →

arxiv 2411.09025 v1 pith:NJWNAG6B submitted 2024-11-13 stat.AP

Jacob Englert , Stefanie Ebelt , Howard Chang This is my paper · ORCID
classification stat.AP MSC 62F1562G0862P10
keywords BayesianadditiveregressiontreessoftBARTnegativebinomialenvironmentalexposuremixturesaccumulatedlocaleffectsasthmaemergencydepartmentvisitsairpollutionAtlanta
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that soft Bayesian additive regression trees can be embedded in a negative binomial count model to estimate smooth, nonlinear, and interacting exposure-response surfaces for mixtures of air pollutants and temperature, with accumulated local effects used to summarize the fitted surface. If the approach works, it gives environmental epidemiologists a practical alternative to Gaussian-process methods for large count datasets. In the Atlanta warm-season data, the mixture model suggests ozone is associated with higher asthma emergency-department visit rates and nitrogen dioxide with lower rates, with both associations stronger on cooler days. The authors are careful to describe these as associations, not causal effects.

What carries the argument

The load-bearing machinery is soft BART as a nonparametric component inside the linear predictor of a negative binomial model, together with Polya-gamma data augmentation that makes the tree-leaf updates conditionally conjugate. Soft trees replace hard decision rules with smooth gating functions, so the ensemble approximates smooth exposure-response surfaces without requiring many trees. A sparsity-inducing Dirichlet prior helps the ensemble ignore noise exposures. Accumulated local effects then summarize the fitted surface by integrating conditional partial derivatives over the observed exposure distribution, isolating the effect of one or two exposures while averaging over plausible values of the others. The combination of the soft-tree ensemble and the ALE summary is what allows nonlinear, interacting mixture effects to be extracted without extrapolation.

What would settle it

Re-run the same 25-tree mixture model on a dataset where the exposure time series are shifted by several weeks relative to the health outcomes; under the null the ALE curves for all five exposures should be flat near zero, and a strongly non-flat NO2 curve would indicate residual confounding or a model artifact.

Watch

Extended reading notes

Core claim

The paper claims that soft BART, placed inside a Poisson-gamma negative binomial regression with a proper conditional autoregressive prior on ZIP-code random intercepts, recovers nonlinear and interactive mixture exposure-response surfaces for count outcomes, and that accumulated local effects then summarize those surfaces without extrapolating to implausible exposure combinations. In 200 simulation repetitions using a benchmark surface with five signal and five noise exposures, soft BART with a sparse branching-process prior showed low bias, near-nominal credible-interval coverage for the exposure function, and low root mean squared error, outperforming rigid-tree BART. Applied to 478,311 asthma emergency-department visits across 128 Atlanta ZIP codes over eight warm seasons, the 25-tree mixture model had much lower WAIC than single-exposure models, and its ALE summary shows a harmful association for ozone and a negative association for nitrogen dioxide, with both effects more pronounced on lower-temperature days.

Load-bearing premise

The Atlanta findings assume that the seasonal and neighborhood adjustments in the model capture every outside factor that could drive both air pollution levels and asthma visits; if any such factor is missed, the estimated ozone and NO2 associations would be biased.

Editorial extensions

If this is right

  • The proposed negative binomial soft-BART sampler scales to roughly 220,000 daily count observations, a size at which Gaussian-process BKMR becomes computationally infeasible.
  • Simulation results indicate that ALE main-effect and pairwise plots recover the true functional forms of important exposures and correctly flatten for noise exposures across the central 95% of each exposure range.
  • WAIC comparisons in the application favor the five-exposure mixture model over any single-exposure model, implying that analyzing pollutants jointly changes the apparent effect of an individual pollutant such as ozone.
  • The Atlanta results imply that pollutant effects vary across the warm season, with the harmful ozone association and the negative NO2 association both stronger when maximum daily temperature is low.
  • A practitioner wanting to estimate mixture effects for count health data can use the provided software instead of BKMR, at the cost of roughly 20 hours of computing for a dataset of this size.

Reading between the lines

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

  • If the negative NO2 association is real, it may reflect atmospheric chemistry in Atlanta, such as ozone scavenging by nitrogen oxides, rather than a protective biological effect of NO2; a natural test would be to stratify the ALE analysis by ozone concentration and see whether the NO2 association reverses.
  • A skeptical check would be a negative-control outcome analysis, such as injury emergency-department visits, with the same exposure mixture; flat ALE curves for the control outcome would increase confidence in the confounding adjustment.
  • The method could be extended to handle lagged exposure windows explicitly by feeding the BART component a structured distributed-lag form, addressing the limitation the paper acknowledges.
  • Because ALE is computed from the fitted ensemble, the same software template should transfer to other count outcomes and other fused exposure products, such as PM2.5 species or wildfire smoke.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a negative binomial regression model with a soft BART component for the exposure-response surface, combined with accumulated local effects (ALE) summaries for interpreting marginal and joint exposure effects in the presence of correlated mixtures. The model is applied to warm-season asthma emergency department visits in Atlanta (2011-2018) with five daily exposures (PM2.5, NO2, O3, CO, maximum temperature), day-of-year splines, a holiday indicator, ZIP-level poverty, and a proper CAR random intercept. A simulation study using the Friedman surface among five active and five noise exposures shows that soft BART with a sparse branching process and T=25 trees recovers the surface with low bias, low RMSE, and credible interval coverage near 95%, and that ALE main and second-order effect estimates match the truth. In the application, the mixture model improves WAIC over single-exposure models, and the ALE summaries show a negative NO2 association, a harmful O3 association, and temperature modification of pollutant effects.

Significance. If the claims hold, the paper offers a practical alternative to BKMR for count outcomes with large sample sizes: soft BART can capture nonlinear and interacting exposure-response surfaces, and ALE provides interpretable summaries without extrapolating to unrealistic mixture profiles. The simulation is well designed, with a known generating function, a realistic correlation structure among active exposures, and evaluation of bias, coverage, and RMSE across ensemble sizes and prior choices. The code is publicly available. The methodological core is credible and useful. However, the headline applied findings about NO2 and O3 in Atlanta rest on strong confounding assumptions that the simulation does not test, and the paper's own 'not causal' caveat does not resolve that limitation for descriptive associations.

major comments (3)
  1. [Section 5, Eq. (3.1)] The applied conclusions—negative association between NO2 and asthma ED visits and stronger temperature modification of NO2 and O3 effects—are identified only under the assumption that the day-of-year spline, holiday indicator, ZIP-level poverty, and pCAR random intercept absorb all relevant temporal and spatial confounding. The simulation in Section 4 cannot support this assumption because its confounders X are drawn independently of exposures Z and spatial effects ν, so the data-generating process contains no unmeasured confounding. In the Atlanta data, plausible omitted factors such as humidity, pollen, healthcare access, and year-specific coding or access trends may be correlated with both the exposure mixture and asthma ED visits, which would bias the ALE estimates, including the negative NO2 association. The Section 6 statement that the analysis is 'not causal in nature' does not remove this bias from the descriptive estimates. Please add a negative-control reanalysis, adjust for additional confounders, or provide a sensitivity analysis bounding the impact of unmeasured confounding; at minimum, the abstract and discussion should temper the applied claims until such evidence is provided.
  2. [Section 3.3 and Section 5] No MCMC convergence diagnostics are reported for either the simulation or the application. The application uses 5,000 burn-in iterations and 1,000 posterior samples after thinning (Section 5), and Table 1 reports credible interval coverage, but without trace plots, R-hat statistics, or effective sample sizes the reader cannot assess whether the reported 95% credible intervals are reliable. This is load-bearing for the coverage claims in the simulation and for the uncertainty bands in Figures 4 and 5. Please report convergence diagnostics for the MCMC chains, or provide a justification for why the chosen run length and thinning are sufficient for the soft BART/NB sampler in Algorithm 1.
  3. [Section 5, Figure 4 and Section 6] The single-exposure and mixture ALE estimates differ substantially: O3 shifts from essentially null to harmful, and NO2 shows a strong negative association only in the mixture model. This pattern suggests sensitivity to mutual adjustment among correlated pollutants and to the model specification, rather than a robust conditional association. The paper's interpretation in Section 6 offers a plausible mechanism (NO2 as an ozone precursor with an inverse relationship to ozone on warmer days), but no analysis is provided to distinguish real mutual adjustment from artifacts of multicollinearity or from omitted confounding. A correlation matrix or variance-inflation summary for the four pollutants and temperature, and a check of the ALE results under different numbers of quantile intervals K, would substantially strengthen the applied claim. Without such evidence, the headline negative NO2 result should be presented as preliminary and model-dependent.
minor comments (4)
  1. [Figure 2 caption] The caption contains a typo: 'Widley applicable information criterion' should be 'Widely applicable information criterion'.
  2. [Section 3.2] The notation θij ∼ Gamma(ξ, exp(ηij)) should explicitly state the shape/scale parameterization used, since Gamma distributions are sometimes parameterized by shape and rate; this would avoid ambiguity in the mean/variance formulas in Appendix A.1.
  3. [Section 3.4] The ALE estimation description could clarify that the lower and upper evaluation points (zlower_k, zupper_k) refer to the boundaries of each quantile interval, and should mention how the choice of K affects the finite-difference approximation and the smoothness of the resulting ALE curves.
  4. [Section 4 and Section 6] The paper positions soft BART as a more feasible alternative to BKMR for large datasets, but no runtime or performance comparison with BKMR is provided. A brief benchmark on a subsample, or even a discussion of the computational scaling, would make the feasibility claim more concrete.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the simulation validates recovery of f against an independent analytic generating function, and the application's ALE summaries are standard interpretations of the fitted model, not independently predicted quantities.

full rationale

The derivation chain is self-contained. The simulation study generates outcomes from a fixed analytic exposure-response surface f(z) = -10 + f0(z)/5 based on Friedman (1991), independent of the soft BART model, and then checks whether the fitted model and its ALE summaries recover that known surface (Table 1, Figures 1, B2, B3). This is a genuine falsifiable test of the estimation procedure, not a fit renamed as a prediction. The application findings are summaries of the fitted Equation (3.1) via the ALE functional defined in Equation (3.4), so they are descriptive interpretations of the model rather than out-of-sample predictions, and the paper does not claim otherwise. The exposure-response surface f is not defined in terms of any fitted constant, and no parameter is tuned to a subset of data and then presented as a prediction of a closely related quantity. Self-citations appear only as background references for Atlanta asthma data, prior air-pollution associations, and the exposure data-fusion product; none of these citations supplies a load-bearing mathematical premise or a uniqueness theorem that forces the paper's conclusions. The limitation that the Atlanta association is observational and potentially confounded is explicitly acknowledged in Section 6 and is a correctness/confounding issue, not a circularity issue.

Assumptions & free parameters 3 free parameters · 7 assumptions · 0 invented entities

The central claim rests on standard probability identities, the exposure measurement model, the confounding adjustment, and the smoothness of the exposure-response surface. No new entities are introduced. The free parameters are user-chosen tuning and summarization settings rather than fitted constants in a derivation.

free parameters (3)
  • Number of BART trees T = 25 in the application; 10, 25, 50, 100 in the simulation
    Ensemble size is a user choice. In Section 5 the authors select T=25 after comparing WAIC and noting larger ensembles give generally similar results, so the reported ALE curves depend on this choice.
  • Day-of-year spline degrees of freedom = 7 per warm season year
    Chosen in Section 5 to model seasonal patterns. Different degrees of freedom could change residual confounding and therefore the estimated exposure associations.
  • ALE quantile intervals K = 40
    Used for all ALE estimates in the simulation and application. The finite-difference approximation of the partial derivative in Equation (3.4) depends on this binning choice.
assumptions (7)
  • standard math Poisson-gamma mixture yields a marginal negative binomial distribution with overdispersion parameter xi.
    Used to justify the likelihood in Equation (3.1) and derived in Appendix A.1.
  • standard math Polya-gamma augmentation converts the negative binomial likelihood into a Gaussian conditional for the latent outcome y*.
    Enables the Gibbs updates for beta, random intercepts, and BART leaf parameters, following Polson et al. (2013) and Pillow and Scott (2012) as used in Appendix A.2.
  • domain assumption Fused exposure estimates from Senthilkumar et al. (2022) provide valid ZIP-level daily PM2.5, NO2, O3, CO, and temperature values.
    All exposure-response estimates inherit measurement error from the fusion model described in Section 2.2.
  • domain assumption The confounder set (holiday indicator, day-of-year spline, ZIP-level poverty, and pCAR random intercept) is sufficient for the reported associations.
    Entered in Section 5 through Equation (3.1). If residual confounding exists, the ALE associations would be biased.
  • domain assumption The exposure-response surface is smooth enough for soft decision trees and for the finite-difference ALE approximation with K=40 quantile intervals.
    Sections 3.1 and 3.4. The simulation supports this under one class of generating functions, but not across all plausible surfaces.
  • domain assumption ZIP-code random intercepts follow a proper conditional autoregressive prior with first-order adjacency matrix W.
    Section 3.3 and Appendix A.4. Misspecification of spatial dependence could affect the estimated exposure surface f(Z).
  • domain assumption Default soft BART priors from Linero and Yang (2018) are adequate for the exposure-response surface without prior sensitivity analysis.
    Section 3.3 states default priors are used. The analysis depends on these defaults being reasonable for this health outcome and exposure correlation structure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Modeling Joint Health Effects of Environmental Exposure Mixtures with Bayesian Additive Regression Trees." pith.science (2026). https://pith.science/paper/NJWNAG6B

@misc{pith2026241109025,
  author       = {Pith},
  title        = {Pith review of: Modeling Joint Health Effects of Environmental Exposure Mixtures with Bayesian Additive Regression Trees},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NJWNAG6B}},
  note         = {Machine review of arXiv:2411.09025}
}
read the original abstract

Studying the association between mixtures of environmental exposures and health outcomes can be challenging due to issues such as correlation among the exposures and non-linearities or interactions in the exposure-response function. For this reason, one common strategy is to fit flexible nonparametric models to capture the true exposure-response surface. However, once such a model is fit, further decisions are required when it comes to summarizing the marginal and joint effects of the mixture on the outcome. In this work, we describe the use of soft Bayesian additive regression trees (BART) to estimate the exposure-risk surface describing the effect of mixtures of chemical air pollutants and temperature on asthma-related emergency department (ED) visits during the warm season in Atlanta, Georgia from 2011-2018. BART is chosen for its ability to handle large datasets and for its flexibility to be incorporated as a single component of a larger model. We then summarize the results using a strategy known as accumulated local effects to extract meaningful insights into the mixture effects on asthma-related morbidity. Notably, we observe negative associations between nitrogen dioxide and asthma ED visits and harmful associations between ozone and asthma ED visits, both of which are particularly strong on lower temperature days.

Figures

Figures reproduced from arXiv: 2411.09025 by the authors.

Figure 1
Figure 1. ALE main effects from the simulation study with [PITH_FULL_IMAGE:figures/full_fig_p013_1.png] view at source ↗
Figure 2
Figure 2. Widley applicable information criterion (WAIC) (Watanabe and Opper, 2010; Gelman [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Estimated relative risk (and 95% posterior credible interval) when all exposures are simultaneously [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Main effect ALE for exposures individually (blue) and as part of a mixture (red) in the Atlanta [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]
Figure 5
Figure 5. Figure 5: Pairwise second-order ALE for each chemical exposure with maximum temperature for the mixture [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 35 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if new.block note output fin.entry FUNCTION b...

  2. [2]

    Agency for Healthcare Research and Quality, Rockville, MD . (2021). HCUPnet , Healthcare Cost and Utilization Project

    work page 2021

  3. [3]

    (2016, March)

    Alhanti, Brooke A, Chang, Howard H, Winquist, Andrea, Mulholland, James A, Darrow, Lyndsey A and Sarnat, Stefanie Ebelt . (2016, March). Ambient air pollution and emergency department visits for asthma: a multi-city assessment of effect modification by age. Journal of Exposure Science & Environmental Epidemiology\/ 26(2), 180--188

    work page 2016

  4. [4]

    (2023, December)

    Antonelli, Joseph, Wilson, Ander and Coull, Brent A . (2023, December). Multiple exposure distributed lag models with variable selection. Biostatistics\/ 25(1), 1--19

    work page 2023

  5. [5]

    and Zhu, Jingyu

    Apley, Daniel W. and Zhu, Jingyu . (2020, September). Visualizing the Effects of Predictor Variables in Black Box Supervised Learning Models . Journal of the Royal Statistical Society Series B: Statistical Methodology\/ 82(4), 1059--1086

    work page 2020

  6. [6]

    and Ebelt, Stefanie

    Bi, Jianzhao, D’Souza, Rohan R., Moss, Shannon, Senthilkumar, Niru, Russell, Armistead G., Scovronick, Noah C., Chang, Howard H. and Ebelt, Stefanie . (2023, April). Acute Effects of Ambient Air Pollution on Asthma Emergency Department Visits in Ten U . S . States . Environmental Health Perspectives\/ 131(4), 047003

    work page 2023

  7. [7]

    (2014, February)

    Bleich, Justin and Kapelner, Adam . (2014, February). Bayesian Additive Regression Trees With Parametric Models of Heteroskedasticity . arXiv:1402.5397 [stat]

    arXiv 2014

  8. [8]

    and Coull, Brent A

    Bobb, Jennifer F., Valeri, Linda, Claus Henn, Birgit, Christiani, David C., Wright, Robert O., Mazumdar, Maitreyi, Godleski, John J. and Coull, Brent A. (2015, July). Bayesian kernel machine regression for estimating the health effects of multi-pollutant mixtures. Biostatistics\/ 16(3), 493--508

    work page 2015

Show all 37 references
  1. [9]

    and Factor-Litvak, Pam

    Carrico, Caroline, Gennings, Chris, Wheeler, David C. and Factor-Litvak, Pam . (2015, March). Characterization of Weighted Quantile Sum Regression for Highly Correlated Data in a Risk Analysis Setting . Journal of Agricultural, Biological, and Environmental Statistics\/ 20(1),...

    2015

  2. [10]

    Centers for Disease Control and Prevention . (2021). Most Recent National Asthma Data

    2021

  3. [11]

    and McCulloch, Robert E

    Chipman, Hugh A., George, Edward I. and McCulloch, Robert E. (2010, March). BART : Bayesian additive regression trees. The Annals of Applied Statistics\/ 4(1), 266--298

    2010

  4. [12]

    (1991, March)

    Friedman, Jerome H. (1991, March). Multivariate Adaptive Regression Splines . The Annals of Statistics\/ 19(1), 1--67

    1991

  5. [13]

    (2001, October)

    Friedman, Jerome H. (2001, October). Greedy function approximation: A gradient boosting machine. The Annals of Statistics\/ 29(5), 1189--1232

    2001

  6. [14]

    (2014, November)

    Gelman, Andrew, Hwang, Jessica and Vehtari, Aki . (2014, November). Understanding predictive information criteria for Bayesian models. Statistics and Computing\/ 24(6), 997--1016

    2014

  7. [15]

    (2000, August)

    Hastie, Trevor and Tibshirani, Robert . (2000, August). Bayesian backfitting (with comments and a rejoinder by the authors. Statistical Science\/ 15(3), 196--223

    2000

  8. [16]

    (2020, March)

    Hill, Jennifer, Linero, Antonio and Murray, Jared . (2020, March). Bayesian Additive Regression Trees : A Review and Look Forward . Annual Review of Statistics and Its Application\/ 7(1), 251--278

    2020

  9. [17]

    Ji, Meng, Cohan, Daniel S and Bell, Michelle L . (2011). Meta-analysis of the Association between Short - Term Exposure to Ambient Ozone and Respiratory Hospital Admissions . Environmental Research Letters\/ 6(2), 024006

    2011

  10. [18]

    (2020, April)

    Keil, Alexander P., Buckley, Jessie P., O’Brien, Katie M., Ferguson, Kelly K., Zhao, Shanshan and White, Alexandra J. (2020, April). A Quantile - Based g- Computation Approach to Addressing the Effects of Exposure Mixtures . Environmental Health Perspectives\/ 128(4), 047004

    2020

  11. [19]

    (2023, July)

    Lappe, Brooke L., Ebelt, Stefanie, D'Souza, Rohan R., Manangan, Arie, Brown, Claudia, Saha, Shubhayu, Harris, Drew, Chang, Howard H., Sole, Adam and Scovronick, Noah . (2023, July). Pollen and asthma morbidity in Atlanta : A 26-year time-series study. Environment International...

    2023

  12. [20]

    (2018, April)

    Linero, Antonio R. (2018, April). Bayesian Regression Trees for High - Dimensional Prediction and Variable Selection . Journal of the American Statistical Association\/ 113(522), 626--636

    2018

  13. [21]

    and Yang, Yun

    Linero, Antonio R. and Yang, Yun . (2018, November). Bayesian Regression Tree Ensembles that Adapt to Smoothness and Sparsity . Journal of the Royal Statistical Society Series B: Statistical Methodology\/ 80(5), 1087--1110

    2018

  14. [22]

    (2023, March)

    Mork, Daniel and Wilson, Ander . (2023, March). Estimating perinatal critical windows of susceptibility to environmental mixtures via structured Bayesian regression tree pairs. Biometrics\/ 79(1), 449--461

    2023

  15. [23]

    (2024, January)

    Mutiso, Fedelis, Li, Hong, Pearce, John L, Benjamin-Neelon, Sara E, Mueller, Noel T and Neelon, Brian . (2024, January). Bayesian kernel machine regression for count data: modelling the association between social vulnerability and COVID -19 deaths in South Carolina . Journal o...

    2024

  16. [24]

    (2017, February)

    O'Lenick, Cassandra R, Winquist, Andrea, Mulholland, James A, Friberg, Mariel D, Chang, Howard H, Kramer, Michael R, Darrow, Lyndsey A and Sarnat, Stefanie Ebelt . (2017, February). Assessment of neighbourhood-level socioeconomic status as a modifier of air pollution–asthma as...

    2017

  17. [25]

    Pillow, Jonathan and Scott, James . (2012). Fully Bayesian inference for neural models with negative-binomial spiking. In: Pereira, F., Burges, C. J., Bottou, L. and Weinberger, K. Q. (editors), Advances in Neural Information Processing Systems \/ , Volume 25. Curran Associates, Inc

    2012

  18. [26]

    and Windle, Jesse

    Polson, Nicholas G., Scott, James G. and Windle, Jesse . (2013, December). Bayesian Inference for Logistic Models Using Pólya – Gamma Latent Variables . Journal of the American Statistical Association\/ 108(504), 1339--1349

    2013

  19. [27]

    (2016, September)

    Pratola, Matthew T. (2016, September). Efficient Metropolis – Hastings Proposal Mechanisms for Bayesian Regression Tree Models . Bayesian Analysis\/ 11(3), 885--911

    2016

  20. [28]

    and Mulholland, James

    Senthilkumar, Niru, Gilfether, Mark, Chang, Howard H., Russell, Armistead G. and Mulholland, James . (2022, April). Using land use variable information and a random forest approach to correct spatial mean bias in fused CMAQ fields for particulate and gas species. Atmospheric E...

    2022

  21. [29]

    Dana, Sarnat, Jeremy A., Waller, Lance A., Sarnat, Stefanie E., Mulholland, James A

    Strickland, Matthew J., Darrow, Lyndsey A., Klein, Mitchel, Flanders, W. Dana, Sarnat, Jeremy A., Waller, Lance A., Sarnat, Stefanie E., Mulholland, James A. and Tolbert, Paige E. (2010, August). Short-term Associations between Ambient Air Pollutants and Pediatric Asthma Emerg...

    2010

  22. [30]

    and Liu, Yang

    Strickland, Matthew J., Hao, Hua, Hu, Xuefei, Chang, Howard H., Darrow, Lyndsey A. and Liu, Yang . (2016, May). Pediatric Emergency Visits and Short - Term Changes in PM _ 2.5 Concentrations in the U . S . State of Georgia . Environmental Health Perspectives\/ 124(5), 690--696...

    2016

  23. [31]

    and Kao, S-C

    Thornton, M.M., Shrestha, R., Wei, Y., Thornton, P.E. and Kao, S-C. (2022). Daymet: Daily Surface Weather Data on a 1-km Grid for North America , Version 4 R1 . Publisher: ORNL Distributed Active Archive Center

    2022

  24. [32]

    Watanabe, Sumio and Opper, Manfred . (2010). Asymptotic Equivalence of Bayes Cross Validation and Widely Applicable Information Criterion in Singular Learning Theory . Journal of Machine Learning Research\/ 11(116), 3571--3594

    2010

  25. [33]

    and Coull, Brent A

    Wilson, Ander, Hsu, Hsiao-Hsien Leon, Chiu, Yueh-Hsiu Mathilda, Wright, Robert O., Wright, Rosalind J. and Coull, Brent A. (2022, June). Kernel machine and distributed lag models for assessing windows of susceptibility to environmental mixtures in children’s health studies. Th...

    2022

  26. [34]

    and Reich, Brian J

    Wilson, Ander, Rappold, Ana G., Neas, Lucas M. and Reich, Brian J. (2014, September). Modeling the effect of temperature on ozone-related mortality. The Annals of Applied Statistics\/ 8(3), 1728--1749

    2014

  27. [35]

    (2014, September)

    Winquist, Andrea, Kirrane, Ellen, Klein, Mitch, Strickland, Matthew, Darrow, Lyndsey A., Sarnat, Stefanie Ebelt, Gass, Katherine, Mulholland, James, Russell, Armistead and Tolbert, Paige . (2014, September). Joint Effects of Ambient Air Pollutants on Pediatric Asthma Emergency...

    2014

  28. [36]

    Zhou, Mingyuan, Li, Lingbo, Dunson, David and Carin, Lawrence . (2012). Lognormal and Gamma Mixed Negative Binomial Regression . Proceedings of the ... International Conference on Machine Learning. International Conference on Machine Learning\/ 2012, 1343--1350

    2012

  29. [37]

    (2012, November)

    İrsoy, Ozan, Yıldız, Olcay Taner and Alpaydın, Ethem . (2012, November). Soft decision trees. In: Proceedings of the 21st International Conference on Pattern Recognition ( ICPR2012 )\/ . pp.\ 1819--1822. Journal Abbreviation: Proceedings of the 21st International Conference on...

    2012

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.