|
Activity Number:
|
217
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #300279 |
|
Title:
|
Power for Genetic Association Studies with Random Allele Frequencies and Genotype Distributions
|
|
Author(s):
|
Walter T. Ambrosius*+ and Ethan Lange and Carl Langefeld
|
|
Companies:
|
Wake Forest University School of Medicine and Wake Forest University School of Medicine and Wake Forest University Health Sciences
|
|
Address:
|
Section on Biostatistics, Winston-Salem, NC, 27157-1063,
|
|
Keywords:
|
power ; genetic association ; qualitative trait ; quantitatitive trait
|
|
Abstract:
|
Estimating the statistical power of a proposed sample size under a specific study design is an important step in planning a study. For a study where assignment to a particular treatment is made, such as in a randomized clinical trial, the calculation of power can be made for many statistical tests using existing methods. This is not the case for many genetic association studies of unrelated individuals where allele and genotype frequencies are random variables. Power calculations for association studies typically ignore this source of variation and are computed assuming Hardy-Weinberg proportions. We propose averaging power over the multinomial distribution to give the true power for a fixed allele frequency and averaging over a Beta prior to account for uncertainty in allele frequency. We present examples for quantitative traits using tests based on the general linear model and for qualitative traits based on Fisher's exact test and the Chi-square test. Our results show that ignoring sampling variability of the multinomial distribution and assuming a fixed allele frequency can lead to overestimates of power.
|
