[ncl-talk] 95% confidence interval

Adam Phillips asphilli at ucar.edu
Wed Feb 6 14:22:47 MST 2019


Hi Kunal,
It looks to me like rtest does not recognize data set to _FillValue. As
that's the case, you can mask your significance array based on the array
coming out of escorc:
ccr = escorc(ts1,ts2)
prob_Dec=rtest(ccr,24,0)
sig_Dec =100*(1-prob_Dec)
sig_Dec = where(ismissing(ccr),sig_Dec at _FillValue,sig_Dec)   <---- set all
points that were missing in ccr to missing in sig_Dec
copy_VarAtts(ts2, sig_Dec)
copy_VarCoords_1(ts1,sig_Dec)

Adam

On Tue, Feb 5, 2019 at 11:33 PM Kunal Bali <kunal.bali9 at gmail.com> wrote:

> Sir, but that shows the confidence level outside the Indian boundary as
> well. how is it possible if the data is only of Indian region and how can
> we remove that?
>
>
>
> ---
> Kunal Bali
>
>
>
>
>
>
> On Wed, Feb 6, 2019 at 1:46 AM Adam Phillips <asphilli at ucar.edu> wrote:
>
>> Hi Kunal,
>> Thank you for sending along an easy-to-run script so one can easily test.
>> I recommended that you take out your mask statements, as there was no need
>> to use those statements (and you were masking all values = 100). Removing
>> those lines showed what the issue is. I altered the script to simply plot
>> the unmasked significance array in raster mode with your 95-100 by 1
>> contours shown, and have attached the results here. As you can see,  most
>> of the plot is equal to 100. The grid cells in purple are less than 95, and
>> thus not considered significant at the 95% level. These areas coincide with
>> your correlation values being near 0 (in the plot you originally sent).
>> Seems reasonable. Thus, as far as the graphics go there is nothing
>> incorrect with what is being plotted. The areas that you thought were
>> greater than 95% were actually less than 95%. Here are the resources I set
>> to get the attached plot:
>>  res2=True
>>   res2 at gsnLeftString     = ""
>>   res2 at gsnRightString     = ""
>>   res2 at gsnDraw              = False           ; Do not draw plot
>>   res2 at gsnFrame             = False           ; Do not advance frome
>>   res2 at cnLinesOn            = True    ; turn off contour lines
>>   res2 at cnLevelSelectionMode = "ManualLevels"   ; manually specify
>> contour levels
>>   res2 at cnMinLevelValF       = 95             ; min level
>>   res2 at cnMaxLevelValF       = 100             ; max level
>>   res2 at cnLevelSpacingF      =  1             ; contour interval
>>   res2 at cnFillOn = True
>>   res2 at cnFillMode = "RasterFill"
>>
>>   plotb=gsn_csm_contour(wks,sig_Dec, res2) ; contours are at 90,100
>>   draw(plotb)
>>   frame(wks)
>>
>> Adam
>>
>> On Mon, Feb 4, 2019 at 10:44 PM Kunal Bali <kunal.bali9 at gmail.com> wrote:
>>
>>> Thanks for the suggestions sir and I modified the script accordingly,
>>> but still, I couldn't able to plot 95% CI.
>>>
>>> I attached the original data with my original script. so please provide
>>> some information on that.
>>>
>>> thanks
>>>
>>>
>>>
>>>
>>> ---
>>> Kunal Bali
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Tue, Feb 5, 2019 at 4:05 AM Adam Phillips <asphilli at ucar.edu> wrote:
>>>
>>>> Hi Kunal,
>>>> I do not have any experience with the rtest function, but I do know
>>>> that you should not need to mask your sig_Dec array. Also note that you are
>>>> creating contours from 90 to 100 by 1, and that you do not really need to
>>>> use gsn_contour_shade.
>>>> Try this:
>>>>
>>>> prob_Dec=rtest(ccr,24,0)
>>>> sig_Dec =100*(1-prob_Dec)
>>>> copy_VarAtts(ts2, sig_Dec)
>>>> copy_VarCoords_1(ts1,sig_Dec)
>>>>
>>>>   res2=True
>>>>   res2 at gsnLeftString     = ""
>>>>   res2 at gsnRightString     = ""
>>>>   res2 at gsnDraw              = False           ; Do not draw plot
>>>>   res2 at gsnFrame             = False           ; Do not advance frome
>>>>   res2 at cnLinesOn            = False    ; turn off contour lines
>>>>   res2 at lbLabelBarOn         = False                            ; turn
>>>> off label bar
>>>>   res2 at cnFillColor          = "black"
>>>>   res2 at cnLevelSelectionMode = "ExplicitLevels"   ; explicitly specify
>>>> contour levels
>>>>   res2 at cnLevels       = 95.            ; set one contour: 95
>>>>   res2 at cnMonoFillPattern = False
>>>>   res2 at cnFillPatterns = (/0,16/)    ; all values .lt.95 set to full
>>>> fill, values .ge.95 =  fill pattern 16
>>>>   res2 at cnMonoFillColor      = False
>>>>   res2 at cnFillColors = (/"transparent","black"/)  ; set all values
>>>> .lt.95 = transparent, all values .ge.95 = black
>>>>
>>>>   plotc=gsn_csm_contour(wks,sig_Dec, res2) ; contour is at 95
>>>>
>>>> If that does not work let ncl-talk know, and send along your sig_Dec
>>>> array (in a .nc file) so everyone can look at the array.
>>>> Adam
>>>>
>>>> On Sat, Feb 2, 2019 at 10:14 AM Kunal Bali <kunal.bali9 at gmail.com>
>>>> wrote:
>>>>
>>>>> Dear NCL users,
>>>>>
>>>>> for computing 95% CI, I used the following code.
>>>>> In that, I am trying to first calculate the 95% CI and then overlay
>>>>> (with pattern/shading fill) with the other plot. But I am not getting the
>>>>> 95% area over the high correlation area. So, could you please let me know
>>>>> whether I am using the correct code or not.
>>>>> The result is also attached herewith.
>>>>>
>>>>>
>>>>> ;**********************************************
>>>>> ;Calculating significance
>>>>> ;**********************************************
>>>>>
>>>>>  prob_Dec=rtest(ccr,24,0)
>>>>>  sig_Dec =100*(1-prob_Dec)
>>>>>
>>>>> printVarSummary(sig_Dec)
>>>>>
>>>>>  copy_VarAtts(ts2, sig_Dec)
>>>>>  copy_VarCoords_1(ts1,sig_Dec)
>>>>>
>>>>>  sig_Dec = mask(sig_Dec, sig_Dec.ge.100,False)
>>>>>  sig_Dec = mask(sig_Dec, sig_Dec.lt.90,False)
>>>>>
>>>>>
>>>>>   res2=True
>>>>>   res2 at gsnLeftString     = ""
>>>>>   res2 at gsnRightString     = ""
>>>>>   res2 at gsnDraw              = False           ; Do not draw plot
>>>>>   res2 at gsnFrame             = False           ; Do not advance frome
>>>>>
>>>>>   ;res2 at cnMonoFillColor      = False
>>>>>   res2 at cnLinesOn            = False    ; turn off contour lines
>>>>>   res2 at lbLabelBarOn         = False                            ; turn
>>>>> off label bar
>>>>>   res2 at cnFillColor          = "black"
>>>>>   res2 at cnLevelSelectionMode = "ManualLevels"   ; manually specify
>>>>> contour levels
>>>>>   res2 at cnMinLevelValF       = 90             ; min level
>>>>>   res2 at cnMaxLevelValF       = 100             ; max level
>>>>>   res2 at cnLevelSpacingF      =  1              ; contour interval
>>>>>   res2 at gsnSpreadColors      = False
>>>>>
>>>>>
>>>>>    plotB=gsn_csm_contour(wks,sig_Dec, res2) ; contours are at 90,100
>>>>>
>>>>>    opt = True
>>>>>    opt at gsnShadeFillType = "pattern" ; pattern fill
>>>>>    opt at gsnShadeHigh =16     ; use pattern #2
>>>>>    opt at gsnShadeFillDotSizeF =  2
>>>>>    plotc=gsn_contour_shade(plotB,-999.,90.,opt)
>>>>>
>>>>>
>>>>>   pmres                        =  True
>>>>>   pmres at gsMarkerColor          = "white"       ;-- marker color
>>>>>   pmres at gsMarkerSizeF          =  50       ;-- set size of marker
>>>>>   pmres at gsLineThicknessF       =  46.          ;-- marker line
>>>>> thickness
>>>>>   pmres at gsMarkerIndex          = 1
>>>>>   pmres at gsFillColor            = "white"
>>>>>
>>>>>
>>>>> SUMMARY:
>>>>>
>>>>> Variable: ts1
>>>>>
>>>>> Type: float
>>>>>
>>>>> Total Size: 389486592 bytes
>>>>>
>>>>>             97371648 values
>>>>>
>>>>> Number of Dimensions: 3
>>>>>
>>>>> Dimensions and sizes: [lat | 2112] x [lon | 1921] x [time | 24]
>>>>>
>>>>> Coordinates:
>>>>>
>>>>>             lat: [5.0078125..37.9921875]
>>>>>
>>>>>             lon: [  68..  98]
>>>>>
>>>>>             time: [9465090..9465113]
>>>>>
>>>>> Number Of Attributes: 8
>>>>>
>>>>>   vmin : -1e+15
>>>>>
>>>>>   vmax : 1e+15
>>>>>
>>>>>   fmissing_value : 1e+15
>>>>>
>>>>>   missing_value : 1e+15
>>>>>
>>>>>   _FillValue : 1e+15
>>>>>
>>>>>   units : kg m-3
>>>>>
>>>>>
>>>>>
>>>>> Variable: ts2
>>>>>
>>>>> Type: float
>>>>>
>>>>> Total Size: 389486592 bytes
>>>>>
>>>>>             97371648 values
>>>>>
>>>>> Number of Dimensions: 3
>>>>>
>>>>> Dimensions and sizes: [lat | 2112] x [lon | 1921] x [time | 24]
>>>>>
>>>>> Coordinates:
>>>>>
>>>>>             lat: [5.0078125..37.9921875]
>>>>>
>>>>>             lon: [  68..  98]
>>>>>
>>>>>             time: [9466530..9466553]
>>>>>
>>>>> Number Of Attributes: 8
>>>>>
>>>>>   vmin : -1e+15
>>>>>
>>>>>   vmax : 1e+15
>>>>>
>>>>>   fmissing_value : 1e+15
>>>>>
>>>>>   missing_value : 1e+15
>>>>>
>>>>>   _FillValue : 1e+15
>>>>>
>>>>>   units : m
>>>>>
>>>>>
>>>>>
>>>>> Variable: ccr
>>>>>
>>>>> Type: float
>>>>>
>>>>> Total Size: 16228608 bytes
>>>>>
>>>>>             4057152 values
>>>>>
>>>>> Number of Dimensions: 2
>>>>>
>>>>> Dimensions and sizes: [2112] x [1921]
>>>>>
>>>>> Coordinates:
>>>>>
>>>>> Number Of Attributes: 1
>>>>>
>>>>>   _FillValue : 1e+15
>>>>>
>>>>> (0) min=-0.996108   max=0.325499
>>>>>
>>>>>
>>>>> Variable: sig_Dec
>>>>>
>>>>> Type: float
>>>>>
>>>>> Total Size: 16228608 bytes
>>>>>
>>>>>             4057152 values
>>>>>
>>>>> Number of Dimensions: 2
>>>>>
>>>>> Dimensions and sizes: [2112] x [1921]
>>>>>
>>>>> Coordinates:
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> ---
>>>>> Kunal
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> ncl-talk mailing list
>>>>> ncl-talk at ucar.edu
>>>>> List instructions, subscriber options, unsubscribe:
>>>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>>>>
>>>>
>>>>
>>>> --
>>>> Adam Phillips
>>>> Associate Scientist,  Climate and Global Dynamics Laboratory, NCAR
>>>> www.cgd.ucar.edu/staff/asphilli/   303-497-1726
>>>>
>>>> <http://www.cgd.ucar.edu/staff/asphilli>
>>>>
>>> _______________________________________________
>>> ncl-talk mailing list
>>> ncl-talk at ucar.edu
>>> List instructions, subscriber options, unsubscribe:
>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>>
>>
>>
>> --
>> Adam Phillips
>> Associate Scientist,  Climate and Global Dynamics Laboratory, NCAR
>> www.cgd.ucar.edu/staff/asphilli/   303-497-1726
>>
>> <http://www.cgd.ucar.edu/staff/asphilli>
>>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk at ucar.edu
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>


-- 
Adam Phillips
Associate Scientist,  Climate and Global Dynamics Laboratory, NCAR
www.cgd.ucar.edu/staff/asphilli/   303-497-1726

<http://www.cgd.ucar.edu/staff/asphilli>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20190206/543f9ba6/attachment.html>


More information about the ncl-talk mailing list