EEG/MEG components: A new statistical approach to analyze their (co)variance properties
Wouter Weeda (NWO PhD student)
Department of Psychology
FMG, University of Amsterdam
Project: NWO project at University of Amsterdam
Project running from: 1 March 2006 – 1 December 2010
Supervisors: Prof. dr M.W. Van der Molen, dr H.M. Huizenga
In this project the primary aim is to assess variance and covariance properties of EEG/MEG components, without the need to localize these components. Such a method should meet several criteria. First, it is necessary that signal variance can be dissociated from noise variance. Second, it should be possible to disentangle latency variance and tests of amplitude and latency variance parameters. Third, it is neccessary that the amplitude covariance between components can be estimated and tested. Existing methods (e.g. variance, complexity, wavelets, independent component analysis, parallel factor analysis) are adequate to answer other reseacrh quetions, but they do not meet the aforementioned criteria, and thus are not suited for the present purposes.
We therefore develop a new statistical method that does meet these criteria. By modeling EEG/MEG by a sum of a) partly random temporal component functions and b) a noise variance model, it will become possible to reliably assess variations in amplitude and latency, and the covariance of amplitudes. Since the proposed method is new and by no means straigthforward, it will be developed in several subprojects that have substantial merits in their own right.
Date of defence: 28 March 2012
Title of thesis: New methods for the analysis of trial-to-trial variability in neuroimaging studies. ISBN: 9789461912121
Promotor: Prof. dr. M.W. van der Molen