Hi Lee,
Billy forwarded your note to me. Could I ask you post this back to the Ferret Users list too? It's always good to close the loop. It's true that it's possible to get a bit tangled up with definitions of variables within a Ferret script.
Ansley
-------- Original Message --------
Begin forwarded message:
> From: lee lindsay <lindsayqd@xxxxxxxxx>
> Date: April 3, 2008 11:42:28 AM PDT
> To: "William.S.Kessler" <William.S.Kessler@xxxxxxxx>
> Subject: Re: [ferret_users] eof_space problem?
>
> Hello, Billy,
>
> I guess sth must be wrong on my script. Before doing the EOF, I
> have done some anomaly analysis and use the anomaly results. Then I
> got the results different with Matlab.
> If I save the anomaly resuts in a new file and redo the EOF, then
> its results are consistent with Matlab results. And also I have
> tried the demo sample case, it's same with the Matlab results. So,
> there's no problem with the Ferret EOF script.
>
> Thanks any way,
> Lindsay
>
>
> 2008/3/30, William.S.Kessler <William.S.Kessler@xxxxxxxx>: Hi
> Lindsay -
>
> That is strange. I have never seen anything like that happen in
> many years of using this program. Can you send me 2 plots:
> eof_space and eof_tfunc for mode 1? Just the 2 plots including all
> the default labels, immediately out of the Ferret definitions.
> Start from a clean ferret session, bring in your data, do the EOF
> definitions, and make the plots. Even better would be to send a 3rd
> plot of a time series at some location (0,140W?) of the original
> data as input to the EOF routines
>
> Billy
>
>
> On 30/03/2008, at 1:31 PM, lee lindsay wrote:
>
> Hello, Billy,
>
> The EOF1 amplitude is reversed, but the TAF1 is same as Matlab
> result, not reversed.
> For example, we have high SST in 1998. By matlab results, the SST
> is high in 1998 and it is reasonalbe. By ferret first space and
> time mode, we got low SST in 1998?
>
> Thanks,
> Lindsay
>
>
> 2008/3/29, William.S.Kessler <William.S.Kessler@xxxxxxxx>: Hi
> Lindsay -
>
> EOF spatial and temporal functions can have either sign, as long as
> the product EOF*Time Amplitude Function has the same sign. That is,
> you can reverse the sign of EOF1, and should find the sign of TAF1
> reversed as well. I think that must be what is happening here. Have
> you checked the Time function against the matlab version? Is it also
> reversed? In that case everything is copacetic. Note that you can
> always multiply the EOF by (-1) (or any constant), as long as you
> divide the TAF by the same constant.
>
> Billy K
>
>
> On 29/03/2008, at 7:47 PM, lee lindsay wrote:
> > Hi, Ansley,
> >
> >
> > I mean the amplitude multiplied by -1 in the first mode only. All
> > the others is same, even the eof_stat shows exactly same percent. I
> > only checked first two modes.
> >
> > My ferret version is 5.7.
> >
> >
> > Here is the matlab eof package.
> > http://hooimeijer-it.com/tudelft/Resources/Matlab_EOFs.html
> >
> > Thanks,
> > Lindsay
> >
> >
> > 2008/3/28, Ansley Manke <Ansley.B.Manke@xxxxxxxx>:
> > Hi Lindsay,
> > We haven't had any reports of this kind of trouble with the EOF
> > functions. When you say the values show inverse, do you mean they
> are
> > reversed north to south? or multiplied by -1?
> >
> > What version of Ferret are you running, and on what computer
> platform?
> >
> > Ansley
> >
> > lee lindsay wrote:
> > >
> > > Hello,
> > >
> > > I have tried to use "eof_space" by ferret and matlab EOF
> method. The
> > > time series results are same by "eof_tfunc" and the second mode is
> > > same. Although the spacial pattern results are similar in the
> first
> > > mode , the values shows inverse. Does anyone know what's wrong on
> > > that? Is it ferret wrong for the first mode spatical pattern?
> > >
> > > Thanks,
> > > Lindsay
> > >
> > >
> >
>
>
>
>