Student Seminar Series - January 25, 2005
University of Minnesota
School of Statistics
College of Liberal Arts
Topics
in Dimension Reduction
Yongwu Shao
Tuesday, January 25, 2005
9:30 AM, 170
Ford Hall
Minneapolis, East Bank Campus
Refreshments at 9:00 AM
300 Ford Hall
Abstract
Dimension Reduction is a useful pre-modeling tool to visualize
and analyze the data. Cook and Weisberg (1991) proposed a method
called Sliced Average Variance Estimator, or SAVE, to estimate the
dimension reduction subspace. Despite the good performance and wide use
of the method, the problem of how to determine the dimension of the
dimension reduction space was not solved until recently when Yin and
Weisberg (2004) proposed a test statistics to do the test. In this talk
we will provide another test statistics and show that it has a simpler
asymptotic distribution than the one got by Yin and Weisberg. We will
also provide a new method to estimate the dimensional reduction space.
This new method requires fewer constraints on the marginal distribution
of the predictors than previous methods and the consistency of it can
be justified by a result of Shen (1997).