Hi Alexander - isn't that part of the calculation in canonical correlation? Often canonical correlation is called a extension of multiple regression in just this way of having a set of dependent variables.
Alexander Sirotkin schrieb: ... > BTW, it occured to me that if S-Plus could do regression > with multiple dependent variables (which is what Connolly & Szalay > method is basically based on) it would be trivial to implement > the gappy PCA algorithm... ??? Is it? Why would that make the implementation of the gappy PCA trivial or even more simple? I'm interested in this idea, but I just can't see the connection... Any hint? Gottfried -- Gottfried Helms Univ. Kassel . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
