|
Activity Number:
|
416
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #309869 |
|
Title:
|
A Simulation Study of the Operating Characteristics of Statistical Analysis Plans for Genome-Wide Association Studies
|
|
Author(s):
|
John Barnard*+
|
|
Companies:
|
The Cleveland Clinic
|
|
Address:
|
Quantitative Health Sciences Wb4, Cleveland, OH, 44195,
|
|
Keywords:
|
multi-locus ; missing data ; power ; multiple comparisons
|
|
Abstract:
|
Genome-wide association (GWA) studies are a potentially powerful approach for detecting genetic associations with complex disease phenotypes. How to best analyze a GWA study remains an open and active area of research. We have conducted numerous simulation studies that attempt to estimate the frequentist operating characteristics of various statistical analysis plans. These plans cover (1) choice of stages; (2) quality control practices; (2) missing data treatment; (4) population stratification approaches; (5) multiple comparison procedures; and (6) choice of test statistics. Simulated data were generated from artificial populations constructed by in-silico random mating starting with a founder population consisting of real GWA data. We will present a summary of the simulation studies and conclusions about recommended statistical analysis plans for GWA studies.
|
