|
Activity Number:
|
61
- Approaches for Modeling Clustered and Longitudinal Data
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #313167
|
|
|
Title:
|
Robust and Scalable Modeling of Intra-Individual Variability in Longitudinal Big Data
|
|
Author(s):
|
Christopher German* and Janet Sinsheimer and Hua Zhou and Jin Zhou
|
|
Companies:
|
and UCLA and University of California, Los Angeles and University of Arizona
|
|
Keywords:
|
Intra-individual variability;
Longitudinal data;
M-estimation
|
|
Abstract:
|
In addition to mean levels, intra-individual variability of longitudinal measurements are risk factors in many health conditions such as cardiovascular disease (blood pressure variability and weight cycling) and diabetes (glycemic control). To create better-targeted interventions, clinicians need to understand what factors influence intra-individual variability. With the increase in mHealth studies collecting data via wearable devices and smartphones compounded with genetic and genomic data as predictors, appropriate statistical methodology must be able to scale to hundreds or thousands of observations per individual. Current methodology for modeling intra-individual variability either employ naïve methods that do not allow for individual-level changes in variability or are computationally infeasible for big data. We utilize a modified linear mixed effects model that allows for simultaneous modeling of mean and intra-individual variability with covariates. We derived an efficient approach to model fitting under an M-estimation framework robust to distributional misspecification of both the response and random effects. We demonstrate our method on simulated and real data examples.
|
Authors who are presenting talks have a * after their name.
