Pith. sign in

REVIEW 4 major objections 5 minor 46 references

Causal Spatial Quantile Regression

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A deep spatial quantile regression estimates distributional treatment effects that vary by location and outcome quantile, using a plug-in estimator of the spatial quantile treatment effect.

desk verdict A solid, incremental extension of SPQR/DeepKriging to spatial quantile treatment effects, with a decent simulation study; the hidden-confounding adjustment is real but less proven than the abstract implies. read the letter →

arxiv 2509.02294 v1 pith:NVINTCWL submitted 2025-09-02 stat.ME

Yan Gong , Reetam Majumder , Brian J. Reich , Raphaël Huser This is my paper · ORCID
classification stat.ME MSC 62G0862G0562M3062P10
keywords spatialquantiletreatmenteffectconfoundingdeepregressionneuralnetworkssemiparametricbirthweightcausalinferenceheterogeneity
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

This paper sets out to show that quantile treatment effects can be estimated as a function of both location and outcome quantile in observational spatial data, even when unobserved confounders are present. It builds a deep spatial quantile regression model in which the conditional outcome distribution is a mixture of monotone spline bases with weights learned by a neural network that sees treatment, covariates, coordinates, and multi-resolution spatial features. The fitted quantiles are plugged into an average difference to estimate the spatial quantile treatment effect (SQTE). The paper's key claim is that this estimator recovers the true treatment effect when hidden confounders are smoother in space than the treatment, because local neighborhood fitting makes such confounders nearly constant. An application to North Carolina birth records finds that maternal smoking lowers birth weight across all quantiles, with the largest harms at the low end of the distribution.

What carries the argument

The central object is the deep spatial quantile regression model: the conditional density of the outcome is written as a weighted mixture of second-order M-spline basis functions, with weights produced by a feed-forward neural network with a softmax output. The network inputs include treatment, covariates, spatial coordinates, and multi-resolution radial-basis spatial features, so quantiles are obtained by inverting the implied spline cumulative distribution function. This gives the plug-in SQTE estimator. The spatial confounding adjustment fits the model on a neighborhood around each target location, using an empirically chosen distance that includes a specified fraction of observations, so

What would settle it

Simulate a hidden confounder with the same spatial range as the treatment, for example by letting treatment propensity depend on a Gaussian process with range equal to that of the confounder, and check whether the SQTE estimator's bias grows. If it does, the claim that the method works in the presence of spatial hidden confounders fails in that regime.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that the plug-in estimator, using quantiles from the deep spatial quantile model, recovers the true location- and quantile-specific treatment effect under the standard potential-outcome assumptions plus a neighborhood spatial confounding adjustment. The adjustment is effective when hidden confounders are smoother in space than the treatment, so local fitting makes them approximately constant and removes their bias. Simulations with no confounders, observed confounders, and hidden spatial and non-spatial confounders support the claim, and the maternal smoking application yields negative spatial quantile treatment effects at all quantiles, with pa

Load-bearing premise

The spatial confounding adjustment works only if hidden confounders are smoother in space than the treatment, so local fitting can treat them as constant; the paper assumes this rather than proves it, and the simulations test only one smooth hidden confounder.

Editorial extensions

If this is right

  • Researchers can quantify causal effects across the whole outcome distribution, not just the mean, in spatial observational data.
  • The estimator identifies especially vulnerable quantiles and regions, such as low birth weight, enabling targeted public-health interventions.
  • Including multi-resolution spatial features reduces prediction error when spatial random effects are present.
  • The neighborhood spatial confounding adjustment reduces residual spatial patterns and bias when hidden confounders are smooth in space.
  • The framework can be ported to other spatially indexed outcomes, such as air pollution, climate, and economic outcomes, because the outcome regression is flexible and quantile-dependent.

Reading between the lines

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

  • Inference: The smoothness premise can be tested in practice by comparing the spatial correlation range of treatment with that of suspected unmeasured confounders; if they are comparable, the neighborhood adjustment cannot be trusted.
  • Inference: The framework should extend to continuous or multi-valued treatments by letting the neural network absorb a treatment-effect surface, though identification would then require a generalized propensity-style assumption.
  • Inference: Because the simulations use a smooth deterministic hidden confounder, the paper does not establish behavior when hidden confounders oscillate at the same spatial scale as treatment; a high-frequency spatial confounder is a natural stress test.
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

4 major / 5 minor

Summary. The paper develops a semiparametric neural-network quantile regression for spatial data, combining DeepKriging spatial basis functions with SPQR monotone spline density estimation. Under the potential-outcomes framework, it defines the spatial quantile treatment effect Δ(τ,s) and proposes a plug-in estimator based on fitted conditional quantiles. A neighborhood-based adjustment is introduced to reduce bias from unobserved spatial confounders. The method is evaluated in simulations with no confounders, observed confounders, and hidden spatial/non-spatial confounders, and is applied to estimate the effect of maternal smoking on birth weight in North Carolina.

Significance. The integration of flexible deep spatial models with distributional causal estimands is a timely contribution, and the birth-weight application is policy-relevant. The simulation study is reasonably extensive (100 replications, 5 models, 3 scenarios) and the results show that the method can estimate SQTE well in the simulated scenarios when spatial structure is captured. However, the central claim about hidden confounders is supported by only one deterministic smooth confounder and no theoretical guarantees are provided; as a result, the current scope claims exceed the evidence. With appropriate qualifications and a clarified adjustment algorithm, the paper could be a useful methodological contribution.

major comments (4)
  1. [§2.3, §3.3.3] The central claim that the neighborhood adjustment removes bias from spatial hidden confounders is supported by a single simulation whose mechanism is not the one described. In §3.3.3, H3(s)=sin(5πs1)+cos(2πs2) and p(s)=expit{5H3(s)}; with a 20% neighborhood, H3 varies substantially across the local window (the sin term changes by >1), so it is not "roughly constant" as required by §2.3. The reported success could be due to Model 5's multi-resolution basis functions or the coordinates rather than local constancy. The abstract's unqualified "even with the presence of spatial hidden confounders" needs either a formal condition on the confounder's smoothness relative to the neighborhood radius or a simulation with a high-frequency confounder.
  2. [§2.2, §5] The plug-in estimator bΔ(τ,sp) has no consistency or asymptotic theory; the paper concedes in §5 that "significant foundational theoretical work is necessary." Because the abstract says the method "can accurately estimate SQTE," the word "accurately" is currently justified only by simulations. Moreover, in the hidden-confounder setting of §3.3.3, Assumption 3 (ignorability) is violated by construction, so it is unclear what population quantity the estimator targets. At minimum, state the sufficient conditions under which bΔ converges and adds a theorem, or replace accuracy claims by simulation-specific statements.
  3. [§3.3.3, Figure 1] The simulation results show that in the hidden-confounder scenario only Model 5 performs as well as the unconfounded case; Models 1–4 have substantially larger RMISE, and the adjustment reduces error only for Models 1 and 2. Thus the blanket statement in the abstract overstates the result. The conclusion should be conditional on including sufficiently rich multi-resolution spatial features (Model 5) and on the adjustment being helpful mainly when spatial structure is otherwise omitted.
  4. [§2.3] The description of the spatial confounding adjustment is not reproducible. It first says "To estimate the SQTE at a specific location sp" but then states that the reference point for the initial subregional model fitting is the center of the whole domain and that the fitted coefficients are used to predict for the whole domain. It is unclear whether one local model is fit near the center and extrapolated, or a separate local model is fit around each target location. Please specify the algorithm precisely (e.g., pseudocode), including how the distance is chosen and how predictions are constructed.
minor comments (5)
  1. [§2.2] The formula for bΔ(τ) contains a stray closing brace: "bΔ(τ ) = 1/P Σ ...{ bΔ(τ, sp)}." Also, the notation np/n_p is ambiguous; clarify that n_p is the number of observations at location p.
  2. [§3.2] The sentence "we specify the number of basis functions to be p = 3^2, 5^2, and 7^2 in the two-dimensional space" renders as "p = 3 2, 52, and 72"; fix the formatting.
  3. [§4] The 95% confidence intervals are plotted in Figure 5, but no method for their construction is stated. Please add details (bootstrap, asymptotic approximation, or other).
  4. [Assumption 1] The label "SUTVA" appears as "SUTV A"; also define the acronym at first use.
  5. [§2.3] The phrase "smoother than the treatment variable in space" is informal. Provide a quantitative definition (e.g., Sobolev smoothness, spatial range, or spectral content) so the condition can be checked in simulations and applications.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular reduction in the derivation; central SQTE estimates are validated against simulated ground truth, but the framework rests on self-cited building blocks and the hidden-confounder adjustment is asserted without a formal guarantee.

full rationale

The paper's central claim is that the plug-in estimator bDelta(tau,s_p) = n_p^{-1}\sum_i [bq_1(tau|X_i,s_p) - bq_0(tau|X_i,s_p)] recovers the SQTE. This is not circular: the estimator is obtained by fitting a flexible conditional quantile model and is evaluated by RMISE against known true SQTE values, such as 2s_1(tau-1/2)^2 in Scenarios 1-3. The simulations use independent ground truth, so the successful estimation is not forced by construction. The spatial confounding adjustment (Section 2.3) is a heuristic: the paper states it is 'particularly effective when the hidden confounders are smoother than the treatment variable in space' but provides no theorem, and the conclusion concedes 'significant foundational theoretical work is necessary to support the proposed semiparametric neural-network-based quantile regression functional class.' These are correctness/robustness limitations, not circular reductions. The citations to DeepKriging, SPQR, and Nychka et al. are to prior modeling machinery; the paper's contribution is combining them with causal SQTE, and no load-bearing argument reduces to an unverified self-citation or to a fitted parameter renamed as a prediction. The concern that H3(s)=sin(5\pi s_1)+cos(2\pi s_2) varies within the 20% neighborhood suggests the reported success may be driven by multi-resolution spatial bases rather than the local-constancy mechanism, but this affects interpretation of a simulation, not circularity. Score 2 reflects non-load-bearing self-citation and an unproven adjustment claim, not a tautological derivation.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on unstated tuning choices (network architecture, spline count, neighborhood radius), the smoothness condition for hidden confounders, and the standard causal assumptions. The response support issue is an unstated modeling fix that must be added for the model to be valid.

free parameters (5)
  • Number of spline basis functions K = not reported
    Controls the granularity of the conditional density and quantile function; the paper never specifies K, yet the approximation quality depends on it.
  • Neural network architecture (depth, width, optimizer, learning rate) = not reported
    The model requires choosing these hyperparameters; the paper only states an L-layer feed-forward network with softmax output.
  • Spatial basis resolutions M and nodes per resolution p = M=1,2,3; p=3^2,5^2,7^2 per level (simulation)
    Used in simulations; scale parameters \(\delta_m\) follow Chen et al. (2024), not derived in this paper.
  • Neighborhood distance for spatial confounding adjustment = 20% of data (simulation); 10%-50% in application sensitivity
    Distance is 'determined empirically' (Section 2.3); the main application distance is not reported, only a sensitivity range.
  • Response transformation parameters (if any) = not reported
    The spline density has support [0,1], so a transformation of Y is needed, but no transformation or its parameters are described.
assumptions (5)
  • domain assumption Assumptions 1-4 (SUTVA, consistency, ignorability, positivity) for potential outcomes
    Needed for identification of SQTE via the plug-in formula; Assumption 3 is acknowledged as likely violated in practice, and the hidden-confounding defense is heuristic rather than formal.
  • domain assumption The conditional density of Y is represented as a finite mixture of second-order M-splines with neural-network weights
    Section 2.1 assumes this representation; no universal approximation theorem is proven for this specific class with space-dependent weights.
  • domain assumption Hidden confounders are smoother in space than the treatment variable
    Section 2.3 states this condition for the local adjustment to remove confounding bias; it is load-bearing for the hidden-confounding claim and untested beyond one smooth simulation setting.
  • ad hoc to paper The response Y has been transformed to [0,1] before fitting the spline density
    The spline mixture has support [0,1], but Y is defined on R in Section 2; the paper never describes such a transformation, so the model as written is not well-defined for arbitrary real Y.
  • domain assumption Multi-resolution Wendland basis functions provide a sufficient spatial representation
    Section 2.1 assumes the spatial features capture spatial structure; the number of resolutions and bandwidths are chosen following Chen et al. (2024), not derived here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Causal Spatial Quantile Regression." pith.science (2026). https://pith.science/paper/NVINTCWL

@misc{pith2026250902294,
  author       = {Pith},
  title        = {Pith review of: Causal Spatial Quantile Regression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NVINTCWL}},
  note         = {Machine review of arXiv:2509.02294}
}
read the original abstract

Treatment effects in a wide range of economic, environmental, and epidemiological applications often vary across space, and understanding the heterogeneity of causal effects across space and outcome quantiles is a critical challenge in spatial causal inference. To effectively capture spatial heterogeneity in distributional treatment effects, we propose a novel semiparametric neural network-based causal framework leveraging deep spatial quantile regression and then construct a plug-in estimator for spatial quantile treatment effects (SQTE). This framework incorporates an efficient adjustment procedure to mitigate the impact of spatial hidden confounders. Extensive simulations across various scenarios demonstrate that our methodology can accurately estimate SQTE, even with the presence of spatial hidden confounders. Additionally, the spatial confounding adjustment procedure effectively reduces neighborhood spatial patterns in the residuals. We apply this method to assess the spatially varying quantile treatment effects of maternal smoking on newborn birth weight in North Carolina, United States. Our findings consistently show negative effects across all birth weight quantiles, with particularly severe impacts observed in the lower quantile regions.

Figures

Figures reproduced from arXiv: 2509.02294 by the authors.

Figure 2
Figure 2. [Scenario 3: With hidden confounders.] Spatial patterns of the mean (of 100 [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. [Scenario 3: With hidden confounders.] Spatial patterns of the mean (of 100 [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Average birth weights (in grams) of newborns from first-time white mothers in [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Estimated QTE (in grams) and 95% confidence intervals from the five models [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Estimated SQTE (in grams) of maternal smoking on the birth weight of newborn [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 42 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter edition editor howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.all := #1 'mid.sentence :=...

    work page

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    and Lieli, R

    Abrevaya, J., Hsu, Y.-C. and Lieli, R. P. (2015) Estimating conditional average treatment effects. Journal of Business & Economic Statistics 33(4), 485--505

    work page 2015

  4. [4]

    and Tomko, M

    Akbari, K., Winter, S. and Tomko, M. (2023) Spatial causality: A systematic review on spatial causal inference. Geographical Analysis 55(1), 56--89

    work page 2023

  5. [5]

    and Lu, L

    Ando, T., Li, K. and Lu, L. (2023) A spatial panel quantile model with unobserved heterogeneity. Journal of Econometrics 232(1), 191--213

    work page 2023

  6. [6]

    and Imbens, G

    Athey, S. and Imbens, G. W. (2015) Machine learning methods for estimating heterogeneous causal effects. Stat 1050(5), 1--26

    work page 2015

  7. [7]

    and Guidolin, M

    Bernardi, M. and Guidolin, M. (2023) The determinants of Airbnb prices in New York City: a spatial quantile regression approach. Journal of the Royal Statistical Society Series C: Applied Statistics 72(1), 104--143

    work page 2023

  8. [8]

    de Bont, J., Krishna, B., Stafoggia, M., Banerjee, T., Dholakia, H., Garg, A., Ingole, V., Jaganathan, S., Kloog, I., Lane, K. et al. (2024) Ambient air pollution and daily mortality in ten cities of India : a causal modelling study. The Lancet Planetary Health 8(7), e433--e440

    work page 2024

Show all 46 references
  1. [9]

    C., Gelfand, A

    Castillo-Mateo, J., As \' n, J., Cebri \'a n, A. C., Gelfand, A. E. and Abaurrea, J. (2023) Spatial quantile autoregression for season within year daily maximum temperature data. The Annals of Applied Statistics 17(3), 2305--2325

    2023

  2. [10]

    Chen, W., Li, Y., Reich, B. J. and Sun, Y. (2024) DeepKriging : Spatially dependent deep neural networks for spatial prediction. Statistica Sinica 34

    2024

  3. [11]

    and Tokdar, S

    Chen, X. and Tokdar, S. T. (2021) Joint quantile regression for spatial data. Journal of the Royal Statistical Society Series B: Statistical Methodology 83(4), 826--852

    2021

  4. [12]

    and Hansen, C

    Chernozhukov, V. and Hansen, C. (2005) An IV model of quantile treatment effects. Econometrica 73(1), 245--261

    2005

  5. [13]

    and Hansen, C

    Chernozhukov, V. and Hansen, C. (2006) Instrumental quantile regression inference for structural and treatment effect models. Journal of Econometrics 132(2), 491--525

    2006

  6. [14]

    L., Neelon, B., Nietert, P

    Davis, M. L., Neelon, B., Nietert, P. J., Hunt, K. J., Burgette, L. F., Lawson, A. B. and Egede, L. E. (2019) Addressing geographic confounding through spatial propensity scores: a study of racial disparities in diabetes. Statistical Methods in Medical Research 28(3), 734--748

    2019

  7. [15]

    and Maathuis, M

    Deuber, D., Li, J., Engelke, S. and Maathuis, M. H. (2024) Estimation and inference of extremal quantile treatment effects for heavy-tailed distributions. Journal of the American Statistical Association 119(547), 2206--2216

    2024

  8. [16]

    (2007) Efficient semiparametric estimation of quantile treatment effects

    Firpo, S. (2007) Efficient semiparametric estimation of quantile treatment effects. Econometrica 75(1), 259--276

    2007

  9. [17]

    and Chen, Z

    Gao, B., Wang, J., Stein, A. and Chen, Z. (2022) Causal inference in spatial statistics. Spatial statistics 50, 100621

    2022

  10. [18]

    and Rappold, A

    Giffin, A., Reich, B., Yang, S. and Rappold, A. (2023) Generalized propensity score approach to causal inference with spatial interference. Biometrics 79(3), 2220--2231

    2023

  11. [19]

    Gilbert, B., Datta, A., Casey, J. A. and Ogburn, E. L. (2021) A causal inference framework for spatial confounding. arXiv preprint arXiv:2112.14946

    2021 arXiv

  12. [20]

    J., Misra, S

    Hitsch, G. J., Misra, S. and Zhang, W. W. (2024) Heterogeneous treatment effects and optimal targeting policy evaluation. Quantitative Marketing and Economics 22(2), 115--168

    2024

  13. [21]

    F., Diggle, P

    Jarner, M. F., Diggle, P. and Chetwynd, A. G. (2002) Estimation of spatial variation in risk using matched case-control data. Biometrical Journal: Journal of Mathematical Methods in Biosciences 44(8), 936--945

    2002

  14. [22]

    Kennedy, E. H. (2023) Towards optimal doubly robust estimation of heterogeneous causal effects. Electronic Journal of Statistics 17(2), 3008--3049

    2023

  15. [23]

    Kennedy, E. H. (2024) Semiparametric doubly robust targeted double machine learning: a review. In Handbook of Statistical Methods for Precision Medicine. Chapman & Hall/CRC. Editors E. Laber, B. Chakraborty, E. E. M. Moodie, T. Cai, and M. van der Laan

    2024

  16. [24]

    M., Steyerberg, E

    Kent, D. M., Steyerberg, E. and Van Klaveren, D. (2018) Personalized evidence based medicine: predictive approaches to heterogeneous treatment effects. BMJ 363

    2018

  17. [25]

    and Wang, H

    Kim, M., Wang, L. and Wang, H. J. (2025) Estimation and Inference of Quantile Spatially Varying Coefficient Models Over Complicated Domains. Journal of the American Statistical Association (just-accepted), 1--23

    2025

  18. [26]

    J., Rose, S

    Van der Laan, M. J., Rose, S. et al. (2011) Targeted Learning: Causal Inference for Observational and Experimental Data. Springer

    2011

  19. [27]

    and Richards, J

    Majumder, R. and Richards, J. (2025) Semi-parametric bulk and tail regression using spline-based neural networks. arXiv preprint arXiv:2504.19994

    2025

  20. [28]

    (1923) Sur les applications de la th \'e orie des probabilit \'e s aux experiences agricoles: Essai des principes

    Neyman, J. (1923) Sur les applications de la th \'e orie des probabilit \'e s aux experiences agricoles: Essai des principes. Roczniki Nauk Rolniczych 10(1), 1--51

    1923

  21. [29]

    and Sain, S

    Nychka, D., Bandyopadhyay, S., Hammerling, D., Lindgren, F. and Sain, S. (2015) A multiresolution Gaussian process model for the analysis of large spatial datasets. Journal of Computational and Graphical Statistics 24(2), 579--599

    2015

  22. [30]

    and Samanta, S

    Papadogeorgou, G. and Samanta, S. (2023) Spatial causal inference in the presence of unmeasured confounding and interference. arXiv preprint arXiv:2303.08218

    2023

  23. [31]

    Reich, B. J. (2012) Spatiotemporal quantile regression for detecting distributional changes in environmental processes. Journal of the Royal Statistical Society Series C: Applied Statistics 61(4), 535--553

    2012

  24. [32]

    J., Fuentes, M

    Reich, B. J., Fuentes, M. and Dunson, D. B. (2011) Bayesian spatial quantile regression. Journal of the American Statistical Association 106(493), 6--20

    2011

  25. [33]

    J., Yang, S., Guan, Y., Giffin, A

    Reich, B. J., Yang, S., Guan, Y., Giffin, A. B., Miller, M. J. and Rappold, A. (2021) A review of spatial causal inference methods for environmental and epidemiological applications. International Statistical Review 89(3), 605--634

    2021

  26. [34]

    o v, J., Semenza, J. C., Dasgupta, S., Robinson, E. J., Abd El Wahed, A., Alcayna, T., Arn \'e s-Sanz, C., Bailey, M., B \

    Rockl \"o v, J., Semenza, J. C., Dasgupta, S., Robinson, E. J., Abd El Wahed, A., Alcayna, T., Arn \'e s-Sanz, C., Bailey, M., B \"a rnighausen, T., Bartumeus, F. et al. (2023) Decision-support tools to build climate resilience against emerging infectious diseases in Europe an...

    2023

  27. [35]

    Rubin, D. B. (1974) Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of Educational Psychology 66(5), 688

    1974

  28. [36]

    Schnell, P. M. and Papadogeorgou, G. (2020) Mitigating unobserved spatial confounding when estimating the effect of supermarket access on cardiovascular disease deaths. Annals of Applied Statistics 14(4)

    2020

  29. [37]

    Sun, S., Moodie, E. E. and Ne s lehov \'a , J. G. (2021) Causal inference for quantile treatment effects. Environmetrics 32(4), e2668

    2021

  30. [38]

    A., Shields, B

    Venkatasubramaniam, A., Mateen, B. A., Shields, B. M., Hattersley, A. T., Jones, A. G., Vollmer, S. J. and Dennis, J. M. (2023) Comparison of causal forest and regression-based approaches to evaluate treatment effect heterogeneity: an application for type 2 diabetes precision ...

    2023

  31. [39]

    and Athey, S

    Wager, S. and Athey, S. (2018) Estimation and inference of heterogeneous treatment effects using random forests. Journal of the American Statistical Association 113(523), 1228--1242

    2018

  32. [40]

    (2019) A closed-form estimator for quantile treatment effects with endogeneity

    W \"u thrich, K. (2019) A closed-form estimator for quantile treatment effects with endogeneity. Journal of Econometrics 210(2), 219--235

    2019

  33. [41]

    Xu, D., Daniels, M. J. and Winterstein, A. G. (2018) A Bayesian nonparametric approach to causal inference on quantiles. Biometrics 74(3), 986--996

    2018

  34. [42]

    G., Majumder, R

    Xu, S. G., Majumder, R. and Reich, B. J. (2022 a ) SPQR : An R Package for Semi-Parametric Density and Quantile Regression. arXiv preprint arXiv:2210.14482

    2022 arXiv

  35. [43]

    Xu, S. G. and Reich, B. J. (2023) Bayesian nonparametric quantile process regression and estimation of marginal quantile effects. Biometrics 79, 151--164

    2023

  36. [44]

    G., Yang, S

    Xu, S. G., Yang, S. and Reich, B. J. (2022 b ) A Bayesian Semiparametric Method For Estimating Causal Quantile Effects. arXiv preprint arXiv:2211.01591

    2022 arXiv

  37. [45]

    and Aral, S

    Yang, J., Eckles, D., Dhillon, P. and Aral, S. (2024) Targeting for long-term outcomes. Management Science 70(6), 3841--3855

    2024

  38. [46]

    J., Canale, A

    Zorzetto, D., Bargagli-Stoffi, F. J., Canale, A. and Dominici, F. (2024) Confounder-dependent Bayesian mixture model: Characterizing heterogeneity of causal effects in air pollution epidemiology. Biometrics 80(2), ujae025

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.