This procedure calculates the principal components of a
dataset. Many IDL routines do this, but I don't really
understand their idiosyncrasies. In particular, the builtin PCOMP
has weird outputs, and chokes whin n_dim >> 1, n_dim << n_data.
This procedure efficiently handles that case.
This procedure is the driver for the PRICOM object class. That
class has methods for projecting new data on to principal
data: An n_dim x n_data array of data points
eval: The eigenvalues associated with each principal component. The
eigenvalues are proportional to the scatter of the data projected
onto the principal component.
evec: The principal components.
mean: The routine subtracts off the mean data vector before
performing the analysis. This keyword holds that mean
June 11 2010: Written by Chris Beaumont