Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 158 - Statistical Methods for High-Dimensional Data in Health Care and Medical Research
Type: Topic-Contributed
Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
Sponsor: Section on Statistics in Epidemiology
Abstract #317617
Title: Empirical Likelihood Approach for Variance Component Test with Repeated Accelerometry Measures
Author(s): Haochang Shou*
Companies: University of Pennsylvania
Keywords: variance component; mixed effects model ; sensor data; empirical likelihood

Wearable sensor data are often collected repeatedly over multiple days in biomedical research. Recently there is an increasing focus on studying the probability densities of daily physical activity measures as a continuous alternative to the conventional physical activity markers derived based on pre-specified categories. Hence there is a need to develop appropriate statistical inference tools for such object data sampled from a bounded metric space. In particular, we focus on investigating the variance component test of the physical activity distributions with repetition. Although the traditional mixed-effects models on scalar outcomes are well studied, the analysis of the variance components for object data with arbitrary distributions remain challenging. We propose a set of empirical likelihood-based methods for the inference of variance component with known and unknown fixed effects. Simulation studies show a better coverage by our proposed methods compared to the commonly used likelihood ratio method with Gaussian assumption. Our proposed tests are then applied to estimate the heritability of physical activity traits using accelerometry data collected in a twin study.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2021 program