Including known covariates can reduce power to detect genetic effects in case-control studies
|Originally posted:||23rd July 2012: 9:18 am|
|Public - anyone can view|
Peter Donnelly and colleagues report an analysis considering the inclusion of non-confounding covariates within genome-wide association studies and provide software that can be used to assess the impact on power within a particular study. They find that, when the disease prevalence is low, including known covariates, such as sex or established genetic associations, can reduce the power to detect new associations.