From Average Effects to Proportions Benefiting: A Tutorial on ETZ-Powered Counterfactual Individual Benefit Analysis
Abstract
Consider the question: "Is it true that at least 80% of patients would benefit more from receiving a new treatment, Rx, than from receiving a standard treatment, C?" Conventional treatment efficacy analyses based on average treatment effects cannot directly answer this question. Yet, even if the true mean treatment effects were known exactly, the answer to this question could still be either "yes" or "no", depending on the variability in patients' treatment responses and the dependence between their potential outcomes. This tutorial introduces a proportion-based perspective on individualized treatment benefit through the ETZ-powered Counterfactual Individual Benefit Analysis framework. We present two metrics, the individual benefit probability and the individual benefit lower bound, which enable confidence-based assessments of whether a prespecified proportion of patients achieves a clinically meaningful treatment benefit. We develop these metrics in both the factual random-pair and counterfactual same-patient settings. In particular, the ETZ decomposition provides a transparent characterization of different sources of outcome variability and enables more precise uncertainty quantification for counterfactual treatment comparisons. We illustrate the framework and its practical implementation using patient-level data from a Phase II randomized clinical trial. By shifting the focus from average treatment effects to population proportions of individual benefit, this tutorial provides a complementary perspective for evaluating treatment efficacy and informing individualized treatment decisions while controlling the type I error rate at a prespecified level.
Related
- lepidoptera: A Domain-Agnostic Engine for Causal Simulation with Mixed Variable Types, Vine Copula Dependence, and Individual-Level Counterfactual Propagation
- Choosing Stimuli Wisely by Using Individual Differences: A Step-by-Step Tutorial for the One-Stimulus-Per-Factor Principle in Vision Science and Optometric Assessment
- Proportions of the non-trivial zeros of the Riemann zeta function
- A Categorical Approach to Euclidean Ratios and Proportions
- FissionGuard: Testing Local Watermark Proportions with FDR Control
BibTeX
@misc{wang2026from,
title = {{From Average Effects to Proportions Benefiting: A Tutorial on ETZ-Powered Counterfactual Individual Benefit Analysis}},
author = {Xingya Wang and Yang Han and Yushi Liu and Szu-Yu Tang and Jason C. Hsu},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.33600},
eprint = {2610.33600},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.33600},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - From Average Effects to Proportions Benefiting: A Tutorial on ETZ-Powered Counterfactual Individual Benefit Analysis
AU - Xingya Wang
AU - Yang Han
AU - Yushi Liu
AU - Szu-Yu Tang
AU - Jason C. Hsu
PY - 2026
DA - 2026/10/24
PB - arXiv
DO - 10.48550/arxiv.2610.33600
UR - https://arxiv.org/abs/2610.33600
AB - Consider the question: "Is it true that at least 80% of patients would benefit more from receiving a new treatment, Rx, than from receiving a standard treatment, C?" Conventional treatment efficacy analyses based on average treatment effects cannot directly answer this question. Yet, even if the true mean treatment effects were known exactly, the answer to this question could still be either "yes" or "no", depending on the variability in patients' treatment responses and the dependence between their potential outcomes. This tutorial introduces a proportion-based perspective on individualized treatment benefit through the ETZ-powered Counterfactual Individual Benefit Analysis framework. We present two metrics, the individual benefit probability and the individual benefit lower bound, which enable confidence-based assessments of whether a prespecified proportion of patients achieves a clinically meaningful treatment benefit. We develop these metrics in both the factual random-pair and counterfactual same-patient settings. In particular, the ETZ decomposition provides a transparent characterization of different sources of outcome variability and enables more precise uncertainty quantification for counterfactual treatment comparisons. We illustrate the framework and its practical implementation using patient-level data from a Phase II randomized clinical trial. By shifting the focus from average treatment effects to population proportions of individual benefit, this tutorial provides a complementary perspective for evaluating treatment efficacy and informing individualized treatment decisions while controlling the type I error rate at a prespecified level.
ER -