Individualized Decision Making Under Partial Identification: ThreePerspectives, Two Optimality Results, and One Paradox

@article{Cui2021IndividualizedDM,
  title={Individualized Decision Making Under Partial Identification: ThreePerspectives, Two Optimality Results, and One Paradox},
  author={Yifan Cui},
  journal={Harvard Data Science Review},
  year={2021}
}
  • Yifan Cui
  • Published 17 June 2021
  • Mathematics, Computer Science
  • Harvard Data Science Review
Unmeasured confounding is a threat to causal inference and gives rise to bi-ased estimates. In this paper, we consider the problem of individualized decision making under partial identification. Firstly, we argue that when faced with unmeasured con-founding, one should pursue individualized decision making using partial identification in a comprehensive manner. We establish a formal link between individualized decision making under partial identification and classical decision theory by… 

Figures and Tables from this paper

Machine Intelligence for Individualized Decision Making Under a Counterfactual World: A Rejoinder
TLDR
This JASA rejoinder provides a novel minimax solution (i.e., a rule that minimizes the maximum regret for so-called opportunists) for individualized decision making/policy assignment.
Safe Policy Learning through Extrapolation: Application to Pre-trial Risk Assessment
TLDR
A robust optimization approach is developed that partially identifies the expected utility of a policy, and then finds an optimal policy by minimizing the worst-case regret, allowing the policy-maker to limit the probability of producing a worse outcome than the existing policy.

References

SHOWING 1-10 OF 54 REFERENCES
Machine Intelligence for Individualized Decision Making Under a Counterfactual World: A Rejoinder
TLDR
This JASA rejoinder provides a novel minimax solution (i.e., a rule that minimizes the maximum regret for so-called opportunists) for individualized decision making/policy assignment.
Rejoinder: Optimal Individualized Decision Rules Using Instrumental Variable Methods
We would like to thank the editors for soliciting a discussion of our article and for giving us the opportunity to respond to the discussants’ insightful and inspiring comments. We are also honored
Confounding-Robust Policy Improvement
TLDR
It is demonstrated that hidden confounding can hinder existing policy learning approaches and lead to unwarranted harm, while the robust approach guarantees safety and focuses on well-evidenced improvement, a necessity for making personalized treatment policies learned from observational data reliable in practice.
Optimal Individualized Decision Rules Using Instrumental Variable Methods
TLDR
Methods to estimate optimal individualized rules and construct asymptotically efficient plug-in estimators of the corresponding average causal effect relative to a prespecified reference rule are developed.
Studies in Resource Allocation Processes: Appendix: An optimality criterion for decision-making under ignorance
TLDR
The axiom systems of these last authors imply the existence of subjective probabilities as weights to be assigned to the different possible states of nature, which provide a foundation for the centuries-old use of probability as a guide to action.
Identification of Causal Effects Using Instrumental Variables
TLDR
It is shown that the instrumental variables (IV) estimand can be embedded within the Rubin Causal Model (RCM) and that under some simple and easily interpretable assumptions, the IV estimand is the average causal effect for a subgroup of units, the compliers.
Optimal Structural Nested Models for Optimal Sequential Decisions
TLDR
A novel “Bayes-frequentist compromise” is proposed that combines honest subjective non- or semiparametric Bayesian inference with good frequentist behavior, even in cases where the model is so large and the likelihood function so complex that standard Bayes procedures have poor frequentist performance.
Bounds on the conditional and average treatment effect with unobserved confounding factors
For observational studies, we study the sensitivity of causal inference when treatment assignments may depend on unobserved confounding factors. We develop a loss minimization approach that
Interval Estimation of Individual-Level Causal Effects Under Unobserved Confounding
TLDR
A functional interval estimator is developed that predicts bounds on the individual causal effects under realistic violations of unconfoundedness and is proved that it converges exactly to the tightest bounds possible on CATE when there may be unobserved confounders.
...
1
2
3
4
5
...