Prediction-Powered Inference

TL;DR

Prediction-powered inference provides a framework for valid statistical inference using machine learning predictions, enabling more accurate and data-efficient conclusions across various scientific fields without assumptions on the prediction algorithms.

Contribution

The paper introduces a new framework for valid statistical inference that leverages machine learning predictions without assumptions, improving confidence interval accuracy and efficiency.

Findings

Valid confidence intervals for various quantities are achievable.

More accurate predictions lead to smaller confidence intervals.

Framework is demonstrated across diverse scientific datasets.

Abstract

Prediction-powered inference is a framework for performing valid statistical inference when an experimental dataset is supplemented with predictions from a machine-learning system. The framework yields simple algorithms for computing provably valid confidence intervals for quantities such as means, quantiles, and linear and logistic regression coefficients, without making any assumptions on the machine-learning algorithm that supplies the predictions. Furthermore, more accurate predictions translate to smaller confidence intervals. Prediction-powered inference could enable researchers to draw valid and more data-efficient conclusions using machine learning. The benefits of prediction-powered inference are demonstrated with datasets from proteomics, astronomy, genomics, remote sensing, census analysis, and ecology.