|
Abstract:
|
EEG is test that measures the electrical activity of human brains. An EEG data set that has motivated our research is concerned with the relation between genetic predisposition and alcoholism. There are two groups of subjects, an alcoholic group of 77 subjects and a control group of 45 subjects. Each subject was exposed to some pictures and the voltage values were measured from 64 sensors placed on the head through 256 time points. We apply Directional Regression (Li and Wang, 2007) to this data to find a dimension reduction space in which two groups are clearly separated by their means and variations. A difficulty is that the dimension of predictor is quite large, which makes it impossible to invert the sample covariance matrix. We combine Directional Regression with Principal Component Analysis as an intermediate reduction step to solve this problem.
|
