Abstract Details
Activity Number:
|
510
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistical Learning and Data Mining
|
Abstract #312594
|
|
Title:
|
Joint Model Selection for Correlated Outcomes
|
Author(s):
|
Yunpeng Zhao*+ and Qing Pan
|
Companies:
|
George Mason University and George Washington University
|
Keywords:
|
High-dimensional learning ;
Correlated outcomes
|
Abstract:
|
Correlated multivariate clinical outcomes often result from one unified underlying disease process. We propose novel lasso type selection algorithm in screening high-dimensional genomic markers for such outcomes jointly. The penalty function depends on both the sizes and the signs of the coefficients for the same candidate marker. Because the directions of the effects from the same marker usually agree with each other for positively correlated outcomes, coefficient vectors whose elements have the same sign will be favored compared to coefficient vectors whose elements points to effects in opposite directions. The penalty function is also designed to achieve fast convergence using a modified local quadratic approximation algorithm. Extensive simulation studies show improved specificity and sensitivity compared to independent lasso procedures for each outcome as well as smaller MSPE. The method is illustrated with a randomized clinical trial in type 1 diabetes patients where SNP markers are selected for three symptoms of microvascular complications in different parts of the body (nephropathy, neuropathy and retinopathy).
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2014 program
|
2014 JSM Online Program Home
For information, contact jsm@amstat.org or phone (888) 231-3473.
If you have questions about the Professional Development program, please contact the Education Department.
The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.
Copyright © American Statistical Association.