Hi, I am performing discriminant analysis (Classical fisher analysis) with very high dimensional data. I do not have enough data to even have a full rank estimation of the covariance matrix. Do you know some relate research on computing eigenvectors of covariance matrices which are not full rank and if there exist some asymptotical properties?
Thanks. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
