Identification and online monitoring of experimental measurement states via Cuscore statistic

TL;DR

This paper introduces a statistical approach using Cuscore and Centred Cuscore statistics for detecting, analyzing, and monitoring state changes in experimental measurements, with applications in nuclear physics experiments for real-time and post-experiment data quality control.

Contribution

The paper develops a novel sequential hypothesis testing method employing Cuscore statistics for online detection and classification of state changes in experimental data.

Findings

Effective detection of state changes in nuclear physics experiments.

Ability to identify change points and classify data segments.

Supports real-time and post-experiment monitoring.

Abstract

We present a statistical method for detecting and analysing state changes in experimental measurements using the Cuscore statistic and its special case, the Centred Cuscore statistic. These statistics are designed to identify deviations in detector responses using sequential hypothesis testing relative to a defined reference state. Applications to charge-changing reaction experiments at the FRagment Separator facility at the GSI Helmholtz Centre for Heavy Ion Research, Germany, and the Second Radioactive Ion Beam Line in Lanzhou at the Institute of Modern Physics, China, demonstrate the ability of these tools to quantify state changes, identify the change point, and classify data segments based on measured states. For long-term online monitoring, we use the exponentially weighted moving average to continuously update computations, enabling the detection of successive changes. This method supports both real-time and post-experiment diagnostics and provides a robust approach for enhancing data integrity and experimental control in nuclear physics and related fields.