Séance Séminaire

Séminaire de Probabilités et Statistique

Heritability estimation in high dimensional linear mixed models

For many complex traits in human population, there is a huge gap between the genetic variance explained by population studies and the variance explained by specific variants found thanks to genome wide association studies (GWAS). To estimate this lacking heritability when the considered trait is the height, Yang et al. (2011) suggested the use of linear mixed models. We propose a novel and efficient methodology to estimate the heritability in high dimensional linear mixed models. Our approach is based on a maximum likelihood strategy and can deal with sparse random effects.We establish that our estimator of the heritability is consistent in the general case under mild assumptions and that it satisfies a central limit theorem in the case where the random effect part is not sparse which gives as a byproduct a confidence interval for the heritability.