BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Alias Detection and Spectral Correction for Locally Stationary Tim
 e Series - Idris Eckley\, Lancaster University
DTSTART:20111125T160000Z
DTEND:20111125T170000Z
UID:TALK32916@talks.cam.ac.uk
CONTACT:Richard Samworth
DESCRIPTION:Aliasing occurs when power exists in a signal at frequencies h
 igher than the\nNyquist rate (which is determined by the sampling rate). W
 hen it occurs\,\naliasing causes high frequency information to wrap round 
 and mimic power at\nlower frequencies. \n\nIt is all too easy to overlook 
 aliasing when conducting an analysis of a\ntime series. Indeed it is rarel
 y tested for\, even though a bispectrum-based\ntest of aliasing for (stati
 onary) time series was proposed by Hinich and\nWolinsky in 1988. For local
 ly stationary series the situation is a bit\ndifferent in that aliasing ca
 n be intermittent\, depending on whether the\nspectrum locally contains fr
 equencies higher than the Nyquist rate or not.\nThis talk will introduce a
  wavelet-based method to separate the spectral\ncomponents of a locally st
 ationary time series into two classes: (i) aliased\nor white noise compone
 nts and (ii) lower frequency uncontaminated\ncomponents. In particular we 
 will consider the case of Shannon wavelets\nwhich can separate components 
 even for signals that are not band-limited.\nFinally\, we show our test wo
 rking on simulated and real data sets. \n\n[Joint work with Guy Nason\, Un
 iversity of Bristol]\n
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR