[PAST EVENT] Mathematics Colloquium and EXTREEMS-QED Lecture: Qin Wang (Virginia Commonwealth University)
Abstract: High dimensional data appear in many research fields, such as marketing, finance, environmental and medical studies. Sufficient dimension reduction is a useful tool to tackle this challenging problem. In this talk, a new formulation is proposed based on the Hellinger integral of order two, introduced as a natural measure of the regression information contained in a predictor subspace. The link between chi-squared divergence and dimension reduction subspaces is the key to our approach, which requires minimal (essentially, just existence) assumptions and is computationally efficient relative to existing methods. Numerical studies and a real data example will be presented.