Classification and testing under robust model assumptions
- In this thesis, we develop methodologies for two well-known problems under more realistic and robust model assumptions. First, we consider the target gene identification problem using gene perturbation data. We propose a simple data-adaptive model by incorporating information across the genome as an alternative to the traditional two-group model testing for differential gene expression. The data-adaptive model is more robust than the traditional two-group model without assuming zero-effect for the null model. Second, we consider a multi-class classification problem where the training and the out-of-sample test data may have different distributions. We propose a method that simultaneously makes good predictions for samples similar to the training data and makes rejections otherwise, as well as methods for performance evaluation in this mismatched setting. The proposed method is robust to distributional changes in the data distribution and has good performance in high dimension.
|Type of resource
|electronic resource; remote; computer; online resource
|1 online resource.
|Wong, Wing Ho
|Wong, Wing Ho
|Degree committee member
|Stanford University, Department of Statistics.
|Statement of responsibility
|Submitted to the Department of Statistics.
|Thesis Ph.D. Stanford University 2019.
- © 2019 by Leying Guan
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...