Functional Data Analysis Seminar - Detailseite

Classical statistics models data in D-dimensional Euclidean space.
However, data often can be assumed to be generated from an underlying physical process such that the ground truth of the observed (discrete) data is a continuous curve or function over a compact domain. Examples are time series of temperature or human gait data or the spatio-temporal responses of blood oxygen usage in a brain while performing a memory task.
As in the past decades time and spatial resolution of measurement devices such as cameras and MRI scanners increased, while their signal-to-noise ratios decreased
due to technological advances, a statistical discipline called functional data analysis (FDA) became increasingly fashionable. The latter assumes that the observation are elements of a function space, in most cases a Hilbert space. But for some inferential and applied task this default option comes under scrutiny.

In this seminar we will introduce in one or two talks basic concepts of FDA and discuss
afterwards for what kind of statistical tasks Hilbert or Banach spaces are to be preferred.

Possible topics include, but are not limited to: Functional depths, outlier detection (graphical
tools like functiboxplots), hypothesis testing and confidence regions, reproducing kernel Hilbert space techniques.

Organization:
The seminar will be a weekly meeting. Each participant is asked to prepare and present
a topic for about 60 minutes. Afterwards we will have a group discussion about the presented work.

This seminar can be a starting point for a Bachelor or Master thesis.

An initial organizational meeting will be probably on the 1st of October, the details will be given
by e-mail if you subscribe to the course. Additionally, it would be great, if you are interested, to
contact me via e-Mail, detailing what is your stochastic/statistic background such that I have an overview
of the level of possible participants and their interest.

An initial organizational meeting will be hold on the 18th of October 11:00-13:00 the first regular meeting of the seminar in the semester.