Abstract:
|
In complex trait mapping, one can expect to gain power by including relevant covariates in the analysis and accounting for any relatedness of sampled individuals. For quantitative trait mapping, variations on linear mixed-model approaches have proven effective at achieving these goals in many contexts. However, application of linear mixed-model approaches to binary traits suffers from power loss when covariate effects are strong. We propose a new framework for binary trait mapping that can be viewed as a hybrid of logistic regression and linear mixed-model approaches. Our method uses a logistic link function which increases power when covariate effects are large. We use an estimating equation and score test approach to hypothesis testing which ensures that the method is computationally rapid for large-scale studies. In terms of power, our method outperforms or performs as well as approaches based on either logistic regression or linear mixed models. The method is applicable in the full spectrum of study designs, ranging from combinations of unrelated individuals and small families to individuals sampled from a complex, inbred pedigree.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.