[Thread Prev][Thread Next][Index]

Re: [ferret_users] regression issue: one variable VS another variables



Yangxing,

I admit this is not elegant at all, but it works:

plot/vs/nolab/over/line {`p[t=@min]`,`p[t=@max]`},{`qhat[t=@min]`,`qhat[t=@max]`}

D.

On Wed, Mar 18, 2009 at 5:41 PM, yangxing zheng <yangxingz@xxxxxxxxx> wrote:
Hi Ansley and David,

  It works. But there is redundant connection when the points are cuddled tightly.

  See the attached figure. The line is not really good even without marker on it.

  yx


--- On Thu, 3/19/09, Ansley Manke <Ansley.B.Manke@xxxxxxxx> wrote:

> From: Ansley Manke <Ansley.B.Manke@xxxxxxxx>
> Subject: Re: [ferret_users] regression issue: one variable VS another variables
> To: "David Wang" <climater@xxxxxxxxx>
> Cc: "yangxing zheng" <yangxingz@xxxxxxxxx>, oar.pmel.ferret_users@xxxxxxxx
> Date: Thursday, March 19, 2009, 5:25 AM
> Hi David,
> Sure, PLOT/VS/LINE  will just connect the dots in the order
> they appear in the variables, with no plot marker.
> Instead of the last line in your example, substitute,
>
>    yes? plot/vs/nolab/over/line p,qhat
>
> Or, to be more accurate, if there's missing data, you
> want to plot p only where there are valid values of P and Q.
>  Look at the regression script
>
>   yes? go/help regresst
>
> and see that the variable pmasked is "p where p and
> q"
>
>   yes? plot/vs/nolab/over/line pmasked,qhat
>
> Ansley
>
> David Wang wrote:
> > Hi Yangxing,
> >
> > Here is an example that you can reproduce:
> >
> > yes? use
> "http://iridl.ldeo.columbia.edu/SOURCES/.Indices/.nino/.NCEP_OIv2/dods"
> > yes? sh d
> >      currently SET data sets:
> >     1>
> http://iridl.ldeo.columbia.edu/SOURCES/.Indices/.nino/.NCEP_OIv2/dods
>  (default)
> >  name     title                             I
> J         K         L
> >  NINO12   NINO12                           ...
> ...       ...       1:328
> >  NINO3    NINO3                            ...
> ...       ...       1:328
> >  NINO34   NINO34                           ...
> ...       ...       1:328
> >  NINO4    NINO4                            ...
> ...       ...       1:328
> >
> > yes? let p = nino3
> > yes? let q = nino34
> > yes? go regresst
> > yes? plot/vs p,q
> > yes? plot/vs/nolab/over/line/sym=dot p,qhat
> >
> > OK, I admit it's not really a straight line, but a
> series of connected dots. Does anyone out there know if
> PLOT/VS can have no marker?
> >
> > D.
> >
> > On Wed, Mar 18, 2009 at 1:19 PM, yangxing zheng
> <yangxingz@xxxxxxxxx
> <mailto:yangxingz@xxxxxxxxx>> wrote:
> >
> >     Hi,
> >
> >      Thanks for help. But I am still not clear.
> >
> >      My question is: sst, a function of time, is an
> independent
> >     variable. and u is a dependent variable, which is
> also a function
> >     of time. How can we make a line plot when u is
> regressed onto sst?
> >     Apparently sst as an independent variable is not
> originally
> >     monotonic, it has be reshaped, right?
> >
> >      The FAQ indicates that regresst only describes
> how to regress a
> >     variable (a function of time) into the independent
> variable T, Am
> >     I right? For example, we can do regression sst vs
> T using regresst.jnl
> >
> >      The attached figure is a scatter plot of u (in
> Y-coordinate) VS
> >     sst ( in X-coordinate). Can you explain more
> details for my case?
> >
> >      yx
> >
> >
> >     --- On Wed, 3/18/09, David Wang
> <climater@xxxxxxxxx
> >     <mailto:climater@xxxxxxxxx>> wrote:
> >
> >     > From: David Wang <climater@xxxxxxxxx
> <mailto:climater@xxxxxxxxx>>
> >     > Subject: Re: [ferret_users] regression issue:
> one variable VS
> >     another variables
> >     > To: "yangxing zheng"
> <yangxingz@xxxxxxxxx
> >     <mailto:yangxingz@xxxxxxxxx>>
> >     > Cc: ferret_users@xxxxxxxx
> <mailto:ferret_users@xxxxxxxx>
> >     > Date: Wednesday, March 18, 2009, 8:29 AM
> >     > Hi Yangxing,
> >     >
> >     > regresst.jnl is what you need. Please follow
> this FAQ:
> >     >
> >     >
> >
> http://ferret.pmel.noaa.gov/FERRET_17sep07/FAQ/analysis/regressions.html
> >     >
> >     > It's the least square fit.
> >     >
> >     > HTH,
> >     > D.
> >     >
> >     > On Tue, Mar 17, 2009 at 7:56 PM, yangxing
> zheng
> >     > <yangxingz@xxxxxxxxx
> <mailto:yangxingz@xxxxxxxxx>> wrote:
> >     >
> >     > > Hi,
> >     > >
> >     > >  I do have two variables like area
> averaged sst, and u
> >     > (zonal winds) from l
> >     > > = 1:100 (for simple case, sst and u only
> has one
> >     > dimension of time)
> >     > >
> >     > >  we can make scatter plot
> >     > >  plot/vs sst,u
> >     > >
> >     > >  Now I want to see how sensitive of u to
> sst, we
> >     > usually plot a line that
> >     > > describe the relationship between u and
> sst. How can
> >     > we make it in ferret?
> >     > >
> >     > >  Attached is one example of scatter
> plot.
> >     > >
> >     > >   Thanks.
> >     > >
> >     > >   yx
> >     > >
> >     > >
> >     > >
> >     >
> >     >
> >     > --
> >     > life grows, death doesn't. there is no
> dress rehearsal.
> >
> >
> >
> >
> >
> >
> > -- life grows, death doesn't. there is no dress
> rehearsal.


     



--
life grows, death doesn't. there is no dress rehearsal.

[Thread Prev][Thread Next][Index]

Contact Us
Dept of Commerce / NOAA / OAR / PMEL / TMAP

Privacy Policy | Disclaimer | Accessibility Statement