Stochastic Tree Ensembles for Estimating Heterogeneous Effects

Nikolay Krantsevich; Jingyu He; P. Richard Hahn

Stochastic Tree Ensembles for Estimating Heterogeneous Effects

Nikolay Krantsevich, Jingyu He, P. Richard Hahn

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

4 Citations (Scopus)

Abstract

Determining subgroups that respond especially well (or poorly) to specific interventions (medical or policy) requires new supervised learning methods tailored specifically for causal inference. Bayesian Causal Forest (BCF) is a recent method that has been documented to perform well on data generating processes with strong confounding of the sort that is plausible in many applications. This paper develops a novel algorithm for fitting the BCF model, which is more efficient than the previous Gibbs sampler. The new algorithm can be used to initialize independent chains of the existing Gibbs sampler leading to better posterior exploration and coverage of the associated interval estimates in simulation studies. The new algorithm is compared to related approaches via simulation studies as well as empirical analysis. © 2023 by the author(s)

Original language	English
Title of host publication	Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023
Editors	Francisco Ruiz, Jennifer Dy, Jan-Willem van de Meent
Publisher	PMLR
Pages	6120-6131
Number of pages	12
Publication status	Published - Apr 2023
Event	The 26th International Conference on Artificial Intelligence and Statistics (AISTATS) - Valencia, Spain Duration: 25 Apr 2023 → 27 Apr 2023

Publication series

Name	Proceedings of Machine Learning Research
Volume	206
ISSN (Print)	2640-3498

Conference

Conference	The 26th International Conference on Artificial Intelligence and Statistics (AISTATS)
Place	Spain
City	Valencia
Period	25/04/23 → 27/04/23

Bibliographical note

Research Unit(s) information for this publication is provided by the author(s) concerned.

Funding

The authors would like to thank the reviewers for the helpful feedback that allowed to improve the readability and organization of the paper. This project was supported by Facebook Statistics for Improving Insights and Decisions research award. Jingyu He gratefully acknowledges the support from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. CityU 21504921).

RGC Funding Information

RGC-funded

Access Output

https://proceedings.mlr.press/v206/krantsevich23a.html

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Krantsevich, N., He, J., & Hahn, P. R. (2023). Stochastic Tree Ensembles for Estimating Heterogeneous Effects. In F. Ruiz, J. Dy, & J.-W. van de Meent (Eds.), Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023 (pp. 6120-6131). (Proceedings of Machine Learning Research; Vol. 206). PMLR. https://proceedings.mlr.press/v206/krantsevich23a.html

@inproceedings{2a0bad7f3a47417b878e16ec62dbb2f5,

title = "Stochastic Tree Ensembles for Estimating Heterogeneous Effects",

abstract = "Determining subgroups that respond especially well (or poorly) to specific interventions (medical or policy) requires new supervised learning methods tailored specifically for causal inference. Bayesian Causal Forest (BCF) is a recent method that has been documented to perform well on data generating processes with strong confounding of the sort that is plausible in many applications. This paper develops a novel algorithm for fitting the BCF model, which is more efficient than the previous Gibbs sampler. The new algorithm can be used to initialize independent chains of the existing Gibbs sampler leading to better posterior exploration and coverage of the associated interval estimates in simulation studies. The new algorithm is compared to related approaches via simulation studies as well as empirical analysis. {\textcopyright} 2023 by the author(s)",

author = "Nikolay Krantsevich and Jingyu He and Hahn, {P. Richard}",

note = "Research Unit(s) information for this publication is provided by the author(s) concerned.; The 26th International Conference on Artificial Intelligence and Statistics (AISTATS) ; Conference date: 25-04-2023 Through 27-04-2023",

year = "2023",

month = apr,

language = "English",

series = "Proceedings of Machine Learning Research",

publisher = "PMLR",

pages = "6120--6131",

editor = "Francisco Ruiz and Jennifer Dy and {van de Meent}, Jan-Willem",

booktitle = "Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023",

}

Krantsevich, N, He, J & Hahn, PR 2023, Stochastic Tree Ensembles for Estimating Heterogeneous Effects. in F Ruiz, J Dy & J-W van de Meent (eds), Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023. Proceedings of Machine Learning Research, vol. 206, PMLR, pp. 6120-6131, The 26th International Conference on Artificial Intelligence and Statistics (AISTATS), Valencia, Spain, 25/04/23. <https://proceedings.mlr.press/v206/krantsevich23a.html>

Stochastic Tree Ensembles for Estimating Heterogeneous Effects. / Krantsevich, Nikolay; He, Jingyu; Hahn, P. Richard.
Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023. ed. / Francisco Ruiz; Jennifer Dy; Jan-Willem van de Meent. PMLR, 2023. p. 6120-6131 (Proceedings of Machine Learning Research; Vol. 206).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - Stochastic Tree Ensembles for Estimating Heterogeneous Effects

AU - Krantsevich, Nikolay

AU - He, Jingyu

AU - Hahn, P. Richard

N1 - Research Unit(s) information for this publication is provided by the author(s) concerned.

PY - 2023/4

Y1 - 2023/4

N2 - Determining subgroups that respond especially well (or poorly) to specific interventions (medical or policy) requires new supervised learning methods tailored specifically for causal inference. Bayesian Causal Forest (BCF) is a recent method that has been documented to perform well on data generating processes with strong confounding of the sort that is plausible in many applications. This paper develops a novel algorithm for fitting the BCF model, which is more efficient than the previous Gibbs sampler. The new algorithm can be used to initialize independent chains of the existing Gibbs sampler leading to better posterior exploration and coverage of the associated interval estimates in simulation studies. The new algorithm is compared to related approaches via simulation studies as well as empirical analysis. © 2023 by the author(s)

AB - Determining subgroups that respond especially well (or poorly) to specific interventions (medical or policy) requires new supervised learning methods tailored specifically for causal inference. Bayesian Causal Forest (BCF) is a recent method that has been documented to perform well on data generating processes with strong confounding of the sort that is plausible in many applications. This paper develops a novel algorithm for fitting the BCF model, which is more efficient than the previous Gibbs sampler. The new algorithm can be used to initialize independent chains of the existing Gibbs sampler leading to better posterior exploration and coverage of the associated interval estimates in simulation studies. The new algorithm is compared to related approaches via simulation studies as well as empirical analysis. © 2023 by the author(s)

UR - http://www.scopus.com/inward/record.url?scp=85165151561&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85165151561&origin=recordpage

M3 - RGC 32 - Refereed conference paper (with host publication)

T3 - Proceedings of Machine Learning Research

SP - 6120

EP - 6131

BT - Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS) 2023

A2 - Ruiz, Francisco

A2 - Dy, Jennifer

A2 - van de Meent, Jan-Willem

PB - PMLR

T2 - The 26th International Conference on Artificial Intelligence and Statistics (AISTATS)

Y2 - 25 April 2023 through 27 April 2023

ER -

Stochastic Tree Ensembles for Estimating Heterogeneous Effects

Abstract

Publication series

Conference

Bibliographical note

Funding

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

ECS: XBART: A Novel Tree-Based Machine Learning Framework for Regression, Classification and Treatment Effect Estimation

Cite this

Stochastic Tree Ensembles for Estimating Heterogeneous Effects

Abstract

Publication series

Conference

Bibliographical note

Funding

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Projects

ECS: XBART: A Novel Tree-Based Machine Learning Framework for Regression, Classification and Treatment Effect Estimation

Cite this