Mike,
Jing has given you part of the answer. I will try to fill up what she
has not mentioned.
I have not tried this, but i have no reason to believe why it should not
work.
1. eofunc_ts should be able to return your time series after varimax
rotation, the same way it does return the time series for the unrotated
eofs. I guess this function regresses (evec) on (data) to obtain the
time series.
2. You can always calculate the percentage variance explained by a
pattern once you have (evec) and its corresponding time series and of
course the data.
a- sum up the variance in the (data) over the same domain that you
had specified to the eof- functions.
b- calculate the variance associated with each (evec) pattern.
To do this multiply evec with the time series, find the variance
and sum it up over the domain
c - (b/a)*100 is the percentage variance associated with each
pattern.
A grads script that can do varimax can be found at
http://mpi-web.dkrz.de/en/misc/software/grads/varimax.gs
saji
Saji N. Hameed, Assistant Researcher,
IPRC-SOEST, 2525 Correa Road, Univ. of Hawaii,
Honolulu, HI 96822, U.S.A
http://iprc.soest.hawaii.edu/~saji
tel.: (808) 956 9534 , fax.: (808) 956 9425
----- Original Message -----
From: ncl-talk-request@ucar.edu
Date: Thursday, March 17, 2005 9:02 am
Subject: ncl-talk Digest, Vol 16, Issue 12
> Send ncl-talk mailing list submissions to
> ncl-talk@ucar.edu
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
> or, via email, send a message with subject or body 'help' to
> ncl-talk-request@ucar.edu
>
> You can reach the person managing the list at
> ncl-talk-owner@ucar.edu
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of ncl-talk digest..."
>
>
> Today's Topics:
>
> 1. Re: ncl-talk Digest, Vol 16, Issue 11 (Mike Notaro)
> 2. Re: Re: ncl-talk Digest, Vol 16, Issue 11 (Jing)
> 3. Re: ncl-talk Digest, Vol 16, Issue 11 (Jing)
>
>
> --------------------------------------------------------------------
> --
>
> Message: 1
> Date: Wed, 16 Mar 2005 13:58:07 -0600
> From: Mike Notaro <mnotaro@wisc.edu>
> Subject: Re: ncl-talk Digest, Vol 16, Issue 11
> To: yangjing@mail.iap.ac.cn, ncl-talk@ucar.edu
> Message-ID: <p06020400be5e3ed6c804@[144.92.130.73]>
> Content-Type: text/plain; charset="us-ascii"
>
> Thanks for your response Jing. Unless I am misreading the website,
> only unrotated eof provides percent variance and time series.
> Under the description of eofunc_varimax, it states that it doesn't
> return the percent variance. But I am not sure about the time series.
> Without the % variance and time series, NCL still can not be used
> for rotated eof.
>
> Does anyone know of a script in matlab/ferret/grads/etc
> that easily performs rotated eof? It is a very important stats
> tool but so hard to find a code to perform it. Particularly, I
> need to be able
> to work around missing data.
>
> I assume the examples on
> http://ngwww.ucar.edu/ngdoc/ng/ref/ncl/functions/eofuncs.html
> are incorrectly using the function "eof", which doesn't exist. I
> assumethey mean "eoffunc".
>
> Thanks,
> Mike
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: http://mailman.ucar.edu/pipermail/ncl-
> talk/attachments/20050316/ed33852f/attachment.htm
> ------------------------------
>
> Message: 2
> Date: Wed, 16 Mar 2005 10:16:03 -1000
> From: Jing <yangjing@mail.iap.ac.cn>
> Subject: Re: Re: ncl-talk Digest, Vol 16, Issue 11
> To: Mike Notaro <mnotaro@wisc.edu>, "ncl-talk@ucar.edu"
> <ncl-talk@ucar.edu>
> Message-ID: <0IDG00HLROAYFU@mail.hawaii.edu>
> Content-Type: text/plain; charset=GB2312
>
> Hi, Mike
>
> 1. the name of "eofunc" is right. but you must intall the Version
> 4.2.0.a032 NCL. then it can work
>
> 2. The routines eofunc and eofunc_ts will work around missing
> values if data@_FillValue is set
>
> Sorry thatI mistake your mail about rotated eof analysis. You are
> right that "Currently, the percent variance explained after the
> rotation is not returned by the function. "
>
> Looking forward to other answers from experts of NCL.
>
>
>
> Best!
>
> Jing
>
>
>
>
>
>
> ------------------------------
>
> Message: 3
> Date: Wed, 16 Mar 2005 09:44:53 -1000
> From: Jing <yangjing@mail.iap.ac.cn>
> Subject: Re: ncl-talk Digest, Vol 16, Issue 11
> To: "mnotaro@wisc.edu" <mnotaro@wisc.edu>
> Cc: "ncl-talk@ucar.edu" <ncl-talk@ucar.edu>
> Message-ID: <0IDG00DEHMUZ7C@mail.hawaii.edu>
> Content-Type: text/plain; charset=GB2312
>
> Hi, Mike Notaro
>
> The new eof function does return you percentage variance you needed
> and what you need in rotated EOF analysis.
>
> Read it carefully, You must be able to find detailed information in
> http://ngwww.ucar.edu/ngdoc/ng/ref/ncl/functions/eofuncs.html
> Especially pay attention to its flexible options and rich resources
> for its output.eg: pcvar (1D float array of length neval)represents
> percent variance associated with each eigenvalue.
>
>
> Good luck
>
>
> Jing
> ***************************************************
> Jing YANG
> Graduate Student
> Internatial Pacific Research Center
> University of Hawaii, Manoa
>
> Phone (o): 808-956-7385
> 1680 East West RD
> Honolulu HI 96822
> United states
>
____________________________________________________________________________________________
> Message: 1
> Date: Tue, 15 Mar 2005 18:51:24 -0600
> From: Mike Notaro <mnotaro@wisc.edu>
> Subject: eofunc_varimax
> To: ncl-talk@ucar.edu
> Message-ID: <p06020400be5d3289e468@[144.92.130.73]>
> Content-Type: text/plain; format=flowed; charset=us-ascii
>
> Hello. I have a question about the new function eofunc_varimax.
> I realize that it still can not provide the % variance explained by
> each eof.
>
> However, how do I get the time series for each
> rotated eof mode? I only know how to get a pattern.
>
> Is there any way to know which is eof1, 2, etc without having the
> % variance explained?
>
> Is there an easy-to-use program available in any computer language
> for computing reof, % var, and time series? I am currently using
> a fortran code wrapped to NCL and it is not ideal.
>
> Thanks,
> Mike
>
> ------------------------------
>
>
>
>
>
>
>
>
> ------------------------------
>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk@ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>
> End of ncl-talk Digest, Vol 16, Issue 12
> ****************************************
>
_______________________________________________
ncl-talk mailing list
ncl-talk@ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
This archive was generated by hypermail 2b29 : Thu Mar 17 2005 - 14:13:12 MST