Vince Calhoun
Department of Psychiatry
Yale University School of Medicine
Title:
Does it take more than two to tango?
The use of higher order statistics for the analysis of functional brain
imaging data.
Abstract:
The use of first and second order statistics is ubiquitous for the
purpose of answering research questions in a statistical framework.
There are some very good reasons for this. The use of higher order
statistics requires more data in order to provide reliable estimates.
Additionally, the assumption of Gaussianity (and thus the use of a
distribution which can be completely characterized by its first two
moments) has yielded many fruitful statistical approaches, the most
common being those based on the general linear model. The advancement of
similar techniques for higher order statistics is more complicated
and often not as widely applicable. However the use of methods involving
higher order statistics can complement traditional approaches and in
some cases can be more intuitive to the application. In this talk we
will discuss the use of higher order statistics in general and with
application to functional magnetic resonance imaging (fMRI) data. In
particular, we will focus upon the emerging use of blind source separation
methods (independent component analysis (ICA)) to reveal
novel information not detected using methods based upon second order
statistics. The talk will be divided into 1) a general introduction to
ICA and some of its statistical properties, 2) a general introduction
to fMRI data, and 3) specific examples from previous and ongoing research
studies comparing model-based and second-order based methods with data-driven
and higher-order based methods. Time permitting; we will
also discuss some simple methods of incorporating prior information
into data-driven approaches.
---------------------------------------------
Seminar to be held in Room 107, 24 Hillhouse at 4:15 pm