报告题目：高维成分数据统计分析理论方法

报告时间：2024-06-12 15:30-17:00

报  告 人：占翔  副教授（北京大学）

报告地点：理学院东北楼四楼报告厅(404)

报告摘要: It is quite common to encounter compositional data in modern data sciences. Most existing statistical methods for compositional data analysis are based on a log-ratio transformation that moves compositional data analysis from simplex to reals. Under this framework, we first investigate novel statistical methods for reliable and reproducible variable selection analysis on compositional predictors. The second part of this talk is about composition-on-composition regression. When both responses and predictors are compositional, the inventory of statistical analysis tools is surprisingly limited. Motivated by data analysis problems with high-dimensional microbiome compositional data, we propose the Composition-On-Composition (COC) regression analysis, which does not require log-ratio transformations and hence can handle excessive zeroes in microbiome data. We introduce a penalized estimation equation approach in COC to improve its estimation accuracy in high-dimensional settings and then establish inference procedures to quantify uncertainties in model estimation and prediction. The proposed methods are evaluated using both numerical simulations and real data applications to demonstrate its validity and superiority.