Re: eofs and current vectors

[William.S.Kessler@xxxxxxxx]

The memory and CPU demands of the eof programs presently implemented 
in Ferret are much more intense because of the capacity to handle 
missing data points. There are nested loops NT^3*NX long to deal with 
gaps. If it is known that the data are complete, and then the problem is
very much easier. If it is also known that NT > NX, then it is simpler
still (use the Numerical Recipes routine svdcmp). Since the Ferret 
routine must be able to deal with all situations, we have a general,
but CPU-intensive routine. 

I have this code implemented in Fortran (basically a front-end to
svdcmp) if anybody wants it, but I am out of the office this week
so you'd have to ask me next Monday.

Billy K