Participant-level quantile surfaces for actigraphy data
提出了一种为每个参与者构建个性化分位数曲面的新方法，用于分析加速度计测量的身体活动数据，同时考虑活动时间和强度，并通过模拟和NHANES数据验证了其优越性。
An innovative approach is introduced for constructing participant-specific quantile surfaces from functional observations. The method captures the complexity of physical activity behaviors measured using accelerometers, accounting for both activity timing and intensity, and provides a comprehensive characterization of each participant’s activity distribution. Functional quantile surface estimation (FQSE) is designed to generate activity surfaces as bivariate functions of time and quantile level, incorporating constraints that ensure non-negativity over time and monotonicity along the quantile dimension. The methodology is evaluated through simulations, demonstrating superior precision and stability compared with recently developed alternatives. Application to accelerometer data from NHANES delivers detailed quantification of activity patterns, capturing both temporal and intensity features.