[disgruntledphd2]

Unsupervised learning is hard, and the pick K problem is probably the hardest part.

For PCA or factor analysis, there's lots of ways but without some way of determining ground truth it's difficult to know if you've done a good job.