Abstract:Uncertainty quantification of causal effects is crucial for safety-critical applications such as personalized medicine. A powerful approach for this is conformal prediction, which has several practical benefits due to model-agnostic finite-sample guarantees. Yet, existing methods for conformal prediction of causal effects are limited to binary/discrete treatments and make highly restrictive assumptions such as known propensity scores. In this work, we provide a novel conformal prediction method for potential outcomes of continuous treatments. We account for the additional uncertainty introduced through propensity estimation so that our conformal prediction intervals are valid even if the propensity score is unknown. Our contributions are three-fold: (1) We derive finite-sample prediction intervals for potential outcomes of continuous treatments. (2) We provide an algorithm for calculating the derived intervals. (3) We demonstrate the effectiveness of the conformal prediction intervals in experiments on synthetic and real-world datasets. To the best of our knowledge, we are the first to propose conformal prediction for continuous treatments when the propensity score is unknown and must be estimated from data.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to quantify the uncertainty of causal effects in the case of continuous treatments (such as drug dosages). Specifically, existing methods have limitations when dealing with continuous variables, mainly focusing on binary or discrete treatments, and usually assume that the propensity score is known. These assumptions are often not valid in practical applications, especially in safety - critical applications such as personalized medicine, where the quantification of the uncertainty of causal effects is crucial. ### Main contributions of the paper: 1. **Proposing a new conformal prediction method**: The authors propose a new conformal prediction method for estimating potential outcomes under continuous treatments. This method can provide effective prediction intervals when the propensity score is unknown. 2. **Algorithm implementation**: An algorithm for calculating these prediction intervals is provided. 3. **Experimental verification**: Through experiments on synthetic data sets and real - world data sets, the effectiveness of the proposed conformal prediction intervals is demonstrated. ### Background and motivation: - **Importance of uncertainty quantification**: In personalized medicine, doctors need to determine treatment plans according to the specific conditions of patients. For example, the selection of chemotherapy dosages requires estimating the response of tumor size to the dosage. Point estimates can only give the average effect, while uncertainty quantification can provide a range to help doctors assess the probability of patient benefit. - **Limitations of existing methods**: Existing causal machine - learning methods usually only focus on point estimates and ignore the uncertainty in prediction. This limits their use in safety - critical applications. - **Advantages of conformal prediction**: Conformal prediction is a powerful uncertainty quantification method, which is model - independent and distribution - free, and can provide prediction intervals for finite samples. ### Challenges: 1. **Distribution changes caused by interventions**: Intervention treatments will change the distribution of covariates, especially the propensity score. This violates the exchangeability assumption in conformal prediction, resulting in the invalidity of standard conformal prediction intervals. 2. **Estimation of the propensity score**: The propensity score is usually unknown and needs to be estimated from the data. This introduces additional uncertainty and needs to be considered when constructing conformal prediction intervals. ### Solutions: - **Conditional calibration**: The authors deal with the distribution changes caused by interventions through conditional calibration to ensure the effectiveness of prediction intervals. - **Solving optimization problems**: By solving optimization problems to estimate the quantiles of non - conformity scores, effective prediction intervals are constructed. ### Experimental results: - **Performance indicators**: The effectiveness of the method is evaluated by calculating whether the empirical coverage of the prediction intervals exceeds the threshold \(1-\alpha\). - **Data sets**: Experiments are carried out using synthetic data sets and real - world data sets to verify the effectiveness of the proposed method. In conclusion, this paper solves the problem of quantifying the uncertainty of causal effects under continuous treatments by proposing a new conformal prediction method, especially the effectiveness and reliability in the case where the propensity score is unknown. This method has important application value in fields such as personalized medicine.

Conformal Prediction for Causal Effects of Continuous Treatments

Conformal Prediction for Causal Effects of Continuous Treatments

Conformal Prediction for Dose-Response Models with Continuous Treatments

Conformal Inference of Counterfactuals and Individual Treatment Effects

Conformal Counterfactual Inference under Hidden Confounding

Conformal causal inference for cluster randomized trials: model-robust inference without asymptotic approximations

Conformal Prediction: A Gentle Introduction

Conformal prediction with conditional guarantees

Conformal Prediction Intervals with Temporal Dependence

Conformal Prediction Under Covariate Shift

Probabilistic Conformal Prediction with Approximate Conditional Validity

Conformal Diffusion Models for Individual Treatment Effect Estimation and Inference

Adaptive Conformal Prediction Intervals Using Data-Dependent Weights With Application to Seismic Response Prediction

Distribution-Free Prediction Intervals Under Covariate Shift, With an Application to Causal Inference

Fair Conformal Predictors for Applications in Medical Imaging

Prediction meets causal inference: the role of treatment in clinical prediction models

Confidence on the Focal: Conformal Prediction with Selection-Conditional Coverage

Adaptive Conformal Inference by Betting

On the Role of Surrogates in Conformal Inference of Individual Causal Effects

Conformal Meta-learners for Predictive Inference of Individual Treatment Effects