Exploratory Analysis of Functional Data
Naveen N. Narisetty
University of Michigan, Ann Arbor

Functional data, where each observation is a function or curve, have become quite common in many applications. Examples include spectrometry curves recorded for a range of wavelengths, tonnage data in stamping processes, noise levels of engines as a function of speed, etc.  “Data depth" is a concept that provides an ordering of the data in terms of how close an observation is to the center of the data cloud. Functional data depth plays an important role in exploratory data analysis of functional data, including identifying central regions, detecting functional outliers, and developing functional box plots. This talk will discuss a new notion of extremal depth and describe its applications.  The new method leads to simultaneous central regions that are closely related to the pointwise bands and thus are very attractive. This is joint work with Prof. Vijay Nair.

2015 QPRC