When: Friday, February 6, 3:00 PM
Where: Tyler 055

Abstract
In this talk, we will first discuss the proposed α-separability for functional data. Functional data consist of random samples observed over a continuum, such as curves over a time range. These data often exhibit two kinds of variation: amplitude variation in the vertical direction and phase variation in the horizontal direction. Separating them in an identifiable manner has been a long-standing challenge in functional data analysis. We introduce the notion of “α-separability” to rigorously address this issue, upon constructing a family of α-indexed metrics on the function space. We demonstrate how the metric-induced Fréchet mean leads to the proposed α-separability and an adjustable model for functional data. The parameter α allows user-defined modeling emphasis between vertical and horizontal features. We showcase our method in several real-data applications. If time permits, we will also discuss the proposed reweighted random forest for predicting health-related outcomes using human microbiome data and demonstrate its applications in the American Gut Project.

Bio
Dr. Tian Wang is a research scientist (official title: staff associate II) in the Department of Biostatistics at Columbia University. He also completed his postdoctoral training in the same department. He received a PhD in Mathematics and a master’s degree in Statistics from Washington University in St. Louis. His current research interests include functional data, omics data, and interdisciplinary collaboration.