Problem
Research questions and friction points this paper is trying to address.
Optimal model selection for CATE estimation
Overcoming non-convexity in ensemble model selection
Doubly robust loss for causal Q-aggregation
Causal Q-Aggregation for CATE Model Selection
🤖 AI Summary
Model selection for conditional average treatment effect (CATE) estimation in personalized decision-making remains challenging due to the fundamental unidentifiability of causal effects. Existing methods rely on strong “model-near-true” assumptions—requiring at least one candidate model to approximate the true CATE well—which are often unrealistic in practice.
Method: We propose causal Q-aggregation, a novel model selection framework built upon a doubly robust loss. Crucially, it operates without any model-near-true assumption.
Contribution/Results: Our method achieves the statistically optimal regret rate of log(M)/n for CATE model selection. Theoretically, we derive the first minimax regret lower bound for CATE model aggregation; extend the framework to settings with instrumental variables and unobserved confounding; and quantify the impact of higher-order remainder terms. Empirically, causal Q-aggregation consistently outperforms state-of-the-art ensemble methods across multiple semi-synthetic benchmarks, offering both strong theoretical guarantees and practical robustness.
📝 Abstract
Accurate estimation of conditional average treatment effects (CATE) is at the core of personalized decision making. While there is a plethora of models for CATE estimation, model selection is a nontrivial task, due to the fundamental problem of causal inference. Recent empirical work provides evidence in favor of proxy loss metrics with double robust properties and in favor of model ensembling. However, theoretical understanding is lacking. Direct application of prior theoretical work leads to suboptimal oracle model selection rates due to the non-convexity of the model selection problem. We provide regret rates for the major existing CATE ensembling approaches and propose a new CATE model ensembling approach based on Q-aggregation using the doubly robust loss. Our main result shows that causal Q-aggregation achieves statistically optimal oracle model selection regret rates of $frac{log(M)}{n}$ (with $M$ models and $n$ samples), with the addition of higher-order estimation error terms related to products of errors in the nuisance functions. Crucially, our regret rate does not require that any of the candidate CATE models be close to the truth. We validate our new method on many semi-synthetic datasets and also provide extensions of our work to CATE model selection with instrumental variables and unobserved confounding.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Uses Q-aggregation for CATE model ensembling
Employs doubly robust loss metrics
Achieves optimal oracle model selection rates
🔎 Similar Papers
No similar papers found.
