204
reads

Lixing Zhu

2011-04-21
10:00:00 - 12:00:00

405 , Mathematics Research Center Building (ori. New Math. Bldg.)

This topic consists of two seminars. The first seminar describes the basic motivation of dimension reduction, some existing methods in this area including the least squares; principal Hessian directions; sliced inverse regression; and sliced average variance estimation. The second seminar introduces its applications in dimension reduction methodology development and variable selection. A discretization-expectation estimation is described, which is again to estimate the central subspace (or effective dimension reduction subspace). Using a response-transformation, a least squares formulation is presented for identifying the index in a general single-index model so that we can investigate an almost necessary and sufficient condition for consistency of variable selection and estimation by the least absolute shrinkage selection operator.