Stats4Grads: Local Polynomial Regression in Higher Dimensions
11 November 2009 14:15 in CM221
Smoothing, and more particularly local polynomial regression, has been widely criticized and ignored in the literature when taken to the multivariate setting. The so-called 'curse of dimensionality' has widely contributed to this, as has the fact that the crucial task of bandwidth selection becomes a lot more difficult in higher dimensions. I have been looking at possible solutions to these problems with the hope that this potentially very useful method of prediction will not be ignored entirely. This talk summarises the work I have done over the last year on this topic, as well as covering a couple of other related areas.
See the Stats4Grads page for more details about this series.