Re: SVD missing

From: Dennis Shea (shea AT cgd.ucar.edu)
Date: Mon Oct 10 2005 - 14:00:30 MDT


I have a question regarding the function svdstd.
>I know it can not handle missing data. I want to
>perform SVD analysis and one of my variables is SST, with
>missing data over the land. If I was to assign zeros to the land
>points at all times (therefore a constant over time),
>would that mess up my SVD results? Otherwise
>I will need to make an array that skips over those
>land points. I do not know enough
>about SVD at this time to know if this is acceptable.

Mike,

I don't think this will work. 'svdstd' standardizes the data
using the standard deviation {...}/std. Any grid point
that is all constant [ie: 0.0] will have a std. drv. of 0.0
which, of course, is not allowed.

Also, note the "Caveats" in the documentation.

D

_______________________________________________
ncl-talk mailing list
ncl-talk@ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk



This archive was generated by hypermail 2b29 : Mon Oct 10 2005 - 20:15:28 MDT