New Methods of Variable Selection and Inference on High Dimensional Data
Sheng Ren · OhioLink ETD Center (Ohio Library and Information Network) · 2017
High dimensional data is everywhere in biomedical informatics research.In this dissertation, our research is focused on variable selection and inference of regression models for high dimensional data.In the first part, we develop a new data-driven simultaneous variable selection and clustering method for high dimensional multinomial regression, with the aim of both solving a kind practical problem and shedding light on the effect of correlation in variable selection.In the second part, we shift our focus to more difficult inference problems, and propose a set of methods to construct model selection confidence set, use it to make valid inference and to quantify uncertainty in model selection and ranking.Simulation studies and analysis of real data examples illustrate the performance and applications of the proposed methods.Writing PhD dissertation is not easy and would not be possible without help from various people.First of all, I would like to express my greatest appreciation to my adviser Dr. Emily L. Kang for her supervision of my PhD research these years.She has given me all her support in helping