High-dimensional statistics, with applications to genome-wide association studies (Q2363092): Difference between revisions

Using high-dimensional linear regression models and Lasso estimation procedures as examples, this paper introduces several important concepts and reviews some fundamental theoretical results of high-dimensional statistics, which include procedures and their statistical properties for regression coefficient estimation, variable screening and selection, and confidence regions and hypothesis testing for single or a group of regression coefficients. The conditions for the major theoretical results are explained in details and sketch of poofs are provided for some important theorems. The procedures are illustrated through data analysis examples from genome-wide association studies. Key references are included in this paper, which makes this paper useful not only to readers who are interested in a quick understanding of basic concepts and recent progresses in high-dimensional statistics but also to those who plan to step into this interesting area of statistical research.