Graph-Based Shrinkage as an Alternative for Linear Mixed Models in Plant Association Studies

• Activity Number: 194
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Wei Liu*+ and Dong Wang
• Companies: University of Nebraska-Lincoln and University of Nebraska-Lincoln
• Keywords: plant association mapping ; lasso ; kinship ; coordinate descent ; graph based shrinkage

Abstract

Traditional linear mixed models have a wide range of applications especially for data from related individuals as in the case for plant association studies. However, estimation and inference with maximum likelihood or restricted maximum likelihood methods can encounter computational difficulties when the number of lines are very large and the population structure is strong. In this study, we explore graph based shrinkage as an alternative to account for the population structure by utilizing fused lasso-type penalty. Preliminary results show that the proposed method can achieve much improved computation speed for large data sets and is competitive for estimation and inference purposes. It can also be extended to a number of other settings.