Anna Louise Schröder
PhD student in the Department of Statistics
Supervised by Prof. Piotr Fryzlewicz
Office: TW1, Room 11.01
· Non-stationary time series
· Multiscale correlation estimation
· Change-point detection in mean and variance
· Change-point detection in time-varying spectra
· Statistical analysis of large data
· Adaptive trend estimation in financial time series via multiscale change-point-induced basis recovery, with Piotr Fryzlewicz. Statistics and Its Interface, 6, 449-461.
change-point detection in epileptic seizure multi-channel EEG data, with
Hernando Ombao from the UCI Space-Time Group. Under review.
The R package FreSpeD (part of our submission) has not been officially released on CRAN. I am happy to provide source code of the submitted package version.
The video clip below illustrates our method’s performance on a live recording of an epileptic seizure. The upper graphic represents electrode channel locations on the human scalp, using the international 10–20 name convention. Channels light up red if a change point is detected in a channel’s normalized spectrum. If the coherence between two channels changes, a lines appears between them. The lower panel shows the recording at one channel, with the red bar indicating the current point in time. At the seizure onset (c. 35s in this visualization) we detect a large number of changes near-simultaneously at channel- and coherence-level. Details on definitions and data are provided upon request.
[Last modified: December 2015 by A. L. Schröder]
Past performance is not a reliable indicator of future results.
© London School of Economics and Political Science 2015