Activity Number:
|
308
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 2, 2016 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Mental Health Statistics Section
|
Abstract #319474
|
|
Title:
|
Depicting Activity Profiles via Multilevel Functional Principal Component Analysis: Association and Prediction
|
Author(s):
|
Jiarui Lu* and Lihong Cui and Kathleen R. Merikangas and Haochang Shou
|
Companies:
|
University of Pennsylvania and National Institute of Mental Health and National Institute of Mental Health and University of Pennsylvania
|
Keywords:
|
Physical Activity ;
MFPCA ;
Hypothesis Tests
|
Abstract:
|
Physical activity is known to be an important phenotype for many diseases and can potentially provide insights to disease pathology via human behaviors. We focus on the data from National Institute of Mental Health family study of affective spectrum disorders to find the group difference in the real-time activity patterns across multiple mental disorders, where the difference was contained mostly in the variations rather than the average. A set of methods was proposed to test the variations across two groups of functions. We first use Multilevel Functional Principal Component Analysis to extract major characteristics of the curves via principle components (PC) and scores (PC scores). We then focus on the PC decompositions for Bipolar type I (BPI) and Major Depressive Disorder (MDD), and conduct hypothesis testings that examine the difference of the two groups. The first method assumes common PCs and test on the resulting PC scores, while the second method test separately on both PCs and scores. Both methods indicate larger variations in the BPI group during night. Such observations can help classifying future patients with available activity measures.
|
Authors who are presenting talks have a * after their name.