Re: significance in correlation plots

From: Mary Haley <haley_at_nyahnyahspammersnyahnyah>
Date: Thu Jul 10 2014 - 15:53:22 MDT

Hi,

I don't think you are calling ShadeLtContour correctly, as it expects five
input arguments and not three. Is this just a typo, or did you get any
errors when you ran your script?

I suggest using gsn_contour_shade in place of ShadeLtContour, which has
been deprecated. Your code would look like this:

opt = True
opt@gsnShadeFillType = "pattern" ; pattern fill
opt@gsnShadeLow = 2 ; use pattern #2
plot2 = gsn_contour_shade(wks,0.05,0.75,opt)

The plots should look the same regardless of which function you use, but
gsn_contour_shade is a bit more versatile.

The two numeric arguments to gsn_contour_shade are the "low" and "high"
values that you want to shade based on. In the example above, then, you
could specify shading below 0.05, between 0.05 and 0.75, and/or above 0.75,
using the gsnShadeLow, gsnShadeMid, and gsnShadeHigh options. If you only
use the Low or High options, then the corresponding high or low value will
just be ignored.

The important thing to note about ShadeLtContour and gsn_contour_shade is
that if you indicate a shade value that falls *between* the contour levels
of your plot, then these routines will only shade below the next closest
contour level that is less than this value. This may be the difference
that you are seeing between the grads and NCL plots?

If you continue to have problems, it would help to see a sample image.

Thanks,

--Mary


On Thu, Jul 10, 2014 at 9:46 AM, Noelia otero <noeli1680@gmail.com> wrote:

> Hi,
>
> I was trying to plot the correlations and adding a contour effect in the
> 95% significant areas. I follow some examples but I am not confident with
> my result. This why I would like to know if I am in the right way to do
> that:
>
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_code.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_csm.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/contributed.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/shea_util.ncl"
>
> begin
>
> a =
> addfile("/home/nof/pruebas/ensem_means/pr_Amon_CanCM4_historical_ensmean_t42_westA_meanJAS.nc","r")
> b =
> addfile("/home/nof/pruebas/WAMI_CEOF/WAMI_Amon_CanCM4_historical_ensmean.nc","r")
>
>
> pr =a->pr
> wami =b->WAMI(:,0,0,0)
> lat = a->lat
> lon = a->lonlat@units="degrees_north"
>
> correl=escorc(wami,pr(lat|:,lon|:,time|:))
> correl!0="lat"
> correl!1="lon"
> correl&lat=lat
> correl&lon=lon
>
> ;***significant values
> sig = 0.05
> um = rtest(correl(lat|:,lon|:),45,0)
> um!0="lat"
> um!1="lon"
> um&lat=lat
> um&lon=lon
>
> wks = gsn_open_wks("pdf","example")
>
> res = True
> res@gsnAddCyclic = False
> res@cnFillOn = True
> res@cnLinesOn = True
> res@mpMinLong = min(correl&lon)
> res@mpMinLatF = min(correl&lat)
> res@mpMaxLonF = max(correl&lon)
> res@mpMaxLatF = max(correl&lat)
> ;plotting first
> plot = gsn_csm_contour_map_ce(wks,correl,res)
>
> res2 = True ; res2 probability plots
> res2@gsnAddCyclic = False
>
> res2@gsnDraw = False ; Do not draw plot
> res2@gsnFrame = False ; Do not advance frome
>
> res2@cnLevelSelectionMode = "ManualLevels" ; set manual contour
> levels
> res2@cnMinLevelValF = 0.00 ; set min contour level
> res2@cnMaxLevelValF = 1.05 ; set max contour level
> res2@cnLevelSpacingF = 0.05 ; set contour spacing
>
> res2@cnInfoLabelOn = False ; turn off info label
>
> res2@cnLinesOn = False ; do not draw contour
> lines
> res2@cnLineLabelsOn = False ; do not draw contour
> labels
>
> res2@cnFillScaleF = 0.6 ; add extra density
> map_ce(wks,correl,res)
> plot2 = gsn_csm_contour(wks,um, res2) ;
> plot2 = ShadeLtContour(plot2, 0.05, 2) ; plotting significant
> areas
>
> overlay (plot, plot2)
>
> draw (plot)
> frame(wks)
>
>
> end
>
> My doubts come from plot2 = ShadeLtContour(plot2, 0.05, 2) , would it be
> OK to show the 95% significance?? because I did the same plot in grads
> (just defining a threshold based on t-student), but the plots looks like a
> little bit different..
>
> Thanks in advance!
>
> Cheers
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>

Received on Thu Jul 10 09:53:22 2014

This archive was generated by hypermail 2.1.8 : Wed Jul 23 2014 - 15:33:47 MDT