Re: EOF: PC normalization

Not quite sure what you want to do.

[1] Please use eofunc_Wrap and eofunc_ts_Wrap.
     The 'cov' have been deprecated. Also, the eofunc
     are faster.

[2] I assume you have weighted the observed anomalies by latitude.

[3] Not sure what you mean by "search for this EOF pattern
     in the dataB set". How are you searching?

[4] Do you want to project (say) an eof pattern derived from A
     onto data set B via regression? If so, Example 4 at
        http://www.ncl.ucar.edu/Applications/eof.shtml
     may be helpful.

[5] I am not sure of normalizing.

On 05/18/2012 01:42 PM, Torben Mueller wrote:
> Dear NCL community,
>
> I have the following problem:
>
> I have two similar data sets (dataA and dataB).
> I calculate an EOF pattern from dataA:
>
> eof = eofcov_Wrap(dataA,neof)
>
> And then search for this EOF pattern in the dataB set, hence getting a
> "pseudo PC":
>
> eof_ts = eofcov_ts_Wrap(dataB,eof)
>
> The question is how to normalize the resulting PC to get zero mean and
> unit variance?
>
> Obviously I can not use the original eigenvalue associated with the
> eof pattern, e.g.:
>
> pcnorm = eof_ts(0,:)/sqrt(eof@eval(0))
>
> As the magnitude of the EOF pattern is different in dataB compared to dataA.
> Does anyone have a suggestion on how to approach this?
>
> Thanks!
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Wed May 23 13:35:30 2012