Michael I. Lichter wrote:
>>If I can continue beating this poor, dead horse just a little longer ...
Back in August, Nick described his method of "disposable principal
component analysis" (see list of steps below), which he concluded with
"discard PC results and proceed with modeling." Clearly, he wouldn't
have done the PCA if it didn't guide his modeling in some way. Does he
use it to determine which variables to retain in his model and which to
discard? Just curious.<<
I suspect that's what Nick meant; hopefully he'll clarify. I often find these kinds of analytic techniques useful in very much the way you describe. So I might not use a PCA or factor score in subsequent modeling but instead use a linear composite or the variables by themselves which are tested in blocks. Factor scores might be optimal in some sense but simple linear composites often have better out of sample performance. PCA is also helpful to screen thru lots of variables for variables that correlate with nothing, likely multicollinearity, etc.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
