Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Causal Inference with Complex Treatments: A Survey

Yingrong Wang, Haoxuan Li, Minqin Zhu, Anpeng Wu, Ruoxuan Xiong, Fei Wu, Kun Kuang

TL;DR

This survey addresses causal inference when treatments are complex (multi-valued, continuous, or bundles) and observational data are used. It organizes methods by treatment type and confounding status, covering propensity-based, representation-based, generative modeling, and instrumental-variable approaches, and it catalogues datasets and open-source codes. The work highlights key challenges in identifiability, interaction effects, and data sparsity, and it outlines promising directions such as transfer learning, data augmentation, and robust representations. Overall, the paper provides a comprehensive taxonomy and practical guidance for applying causal methods to real-world, complex treatments across domains like healthcare, economics, and social sciences.

Abstract

Causal inference plays an important role in explanatory analysis and decision making across various fields like statistics, marketing, health care, and education. Its main task is to estimate treatment effects and make intervention policies. Traditionally, most of the previous works typically focus on the binary treatment setting that there is only one treatment for a unit to adopt or not. However, in practice, the treatment can be much more complex, encompassing multi-valued, continuous, or bundle options. In this paper, we refer to these as complex treatments and systematically and comprehensively review the causal inference methods for addressing them. First, we formally revisit the problem definition, the basic assumptions, and their possible variations under specific conditions. Second, we sequentially review the related methods for multi-valued, continuous, and bundled treatment settings. In each situation, we tentatively divide the methods into two categories: those conforming to the unconfoundedness assumption and those violating it. Subsequently, we discuss the available datasets and open-source codes. Finally, we provide a brief summary of these works and suggest potential directions for future research.

Causal Inference with Complex Treatments: A Survey

TL;DR

This survey addresses causal inference when treatments are complex (multi-valued, continuous, or bundles) and observational data are used. It organizes methods by treatment type and confounding status, covering propensity-based, representation-based, generative modeling, and instrumental-variable approaches, and it catalogues datasets and open-source codes. The work highlights key challenges in identifiability, interaction effects, and data sparsity, and it outlines promising directions such as transfer learning, data augmentation, and robust representations. Overall, the paper provides a comprehensive taxonomy and practical guidance for applying causal methods to real-world, complex treatments across domains like healthcare, economics, and social sciences.

Abstract

Causal inference plays an important role in explanatory analysis and decision making across various fields like statistics, marketing, health care, and education. Its main task is to estimate treatment effects and make intervention policies. Traditionally, most of the previous works typically focus on the binary treatment setting that there is only one treatment for a unit to adopt or not. However, in practice, the treatment can be much more complex, encompassing multi-valued, continuous, or bundle options. In this paper, we refer to these as complex treatments and systematically and comprehensively review the causal inference methods for addressing them. First, we formally revisit the problem definition, the basic assumptions, and their possible variations under specific conditions. Second, we sequentially review the related methods for multi-valued, continuous, and bundled treatment settings. In each situation, we tentatively divide the methods into two categories: those conforming to the unconfoundedness assumption and those violating it. Subsequently, we discuss the available datasets and open-source codes. Finally, we provide a brief summary of these works and suggest potential directions for future research.
Paper Structure (54 sections, 65 equations, 8 figures, 4 tables)

This paper contains 54 sections, 65 equations, 8 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Preliminary
  3. Binary Treatment
  4. Under Unconfoundedness
  5. Propensity score (PS)-based Methods
  6. Doubly Robust Methods
  7. Covariate Balancing Methods
  8. Tree-based Methods
  9. Representation-based Methods
  10. Generative Modeling Methods
  11. With Unobserved Confounders
  12. Proxy Variable
  13. Instrumental Variable
  14. Conclusion and Discussion
  15. Multi-valued Treatment
  16. ...and 39 more sections

Figures (8)

  • Figure 1: An example of complex treatments.
  • Figure 2: Potential outcome frameworks w/o unobserved confounders (marked in shadow).
  • Figure 3: Outline of the survey.
  • Figure 4: Categorization of multi-valued treatment methods without unobserved confounders.
  • Figure 5: Categorization of continuous treatment methods without unobserved confounders.
  • ...and 3 more figures