Menu Content/Inhalt
Seminars Print
previous year previous month next month next year
See by year See by month See by week See Today Search Jump to month
Marina Knight, Bristol U. Print
Friday, 30 March 2012, 14:30 - 15:30

Marina Knight, Bristol University


Spectral estimation for non-stationary time series with missing observations

Abstract: Time series that arise in practice often have an inherently irregular sampling structure or missing values. Spectral estimation is a valuable tool in understanding the process behaviour to the point of forecasting it. However, existing methods often assume a regularly-sampled time series, or require 'adaptations' to cope with missing data. Additionally, many techniques also assume that the time series have a stationary behaviour, which in many fields is inappropriate. A new spectral estimation technique for time series that exhibit both a non-stationary behaviour and missing observations is proposed. Instrumental to this new approach are second generation wavelet constructions and a new periodogram is proposed. Theoretical properties of this periodogram are investigated and a bias-corrected estimate is proposed by adopting a penalized least squares criterion.

Location: Plaine, building NO, 9th floor, salle des Professeurs
Contact: Jacqueline Bottemanne, This e-mail address is being protected from spam bots, you need JavaScript enabled to view it