Evaluating Device-Agnostic Properties of Raw Data Preprocessing Metrics in Physical Activity Assessment
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Accelerometry is widely used in physical activity research, but processing approaches differ across studies. This study evaluated the agreement among raw use preprocessing in the entire document metrics (Activity Counts, Activity Index, Euclidean Norm Minus One – ENMO, Mean Amplitude Deviation – MAD, and Monitor Independent Summary Unit – MIMS) derived from two devices. Participants wore ActiGraph (wGT3X‑BT) and Google Pixel Watch 2 devices (50–100 Hz) on their non‑dominant wrist for 24 h. Data were converted to minimum acceleration of the most active minutes (MX) for each metric across 1–60‑s epochs. Agreement between device‑paired MX values was assessed using mean absolute percent error (MAPE) and Bland – Altman analysis. Data from 16 participants (25.4 ± 8.9 years) showed strong agreement for MAD and MIMS MX with MAPE of 6.8% ± 7.0% and 7.1% ± 3.5%. Only MIMS showed lowest bias and consistent MAPE across epochs without compounded error, supporting its use for device‑agnostic assessment. Larger samples and additional device configurations may strengthen generalizability.
Fields of Study & Subjects
Physical Activity and Health
Context-Aware Activity Recognition Systems
Mobile Health and mHealth Applications
Publication details
- Journal / source
- Measurement in Physical Education and Exercise Science
- Pages
- 1–13
- Publication date
- 31 March 2026
- Language
- EN