[ncl-talk] Question about regressing a time series and a gridded data using regCoef_n

[Lyndon Mark Olaguera]

Dear Sir Dennis and Sir Will,

Many thanks for the help!
I got it working already.


Sincerely,

Lyndz

On Thu, Jun 22, 2017 at 10:56 PM, Dennis Shea <shea at ucar.edu> wrote:

> [SNIP]
>  dimsst = dimsizes(sst)
>  ntim   = dimsst(0)
>  nlat   = dimsst(1)
>  mlon   = dimsst(2)
>
> ;****************************************************
> ;Calculate Regression coefficient via regCoef_n
> ;****************************************************
>
>  rc = regCoef_n(data,sst,0,0)
>  printVarSummary(rc)       ;(nlat,mlon)
>                            ;[1D arrays]: nptxy,rstd,yintercept,tval
>
> ;;Variable: rc
> ;;Type: float
> ;;Total Size: 259200 bytes
> ;;            64800 values
> ;;Number of Dimensions: 2
> ;;Dimensions and sizes: [180] x [360]
> ;;Coordinates:
> ;;Number Of Attributes: 5
> ;;  _FillValue :        -1e+30
> ;;  nptxy :     <ARRAY of 64800 elements>
> ;;  rstd :      <ARRAY of 64800 elements>
> ;;  yintercept :        <ARRAY of 64800 elements>
> ;;  tval :      <ARRAY of 64800 elements>
>
> ;****************************************************
> ;Assign coordinates
> ;****************************************************
>  rc!0 = "lat"
>  rc!1 = "lon"
>  rc&lat = sst&lat
>  rc&lon = sst&lon
>
> ;****************************************************
> ;Test for significance
> ;****************************************************
>  df   = onedtond(rc at nptxy-2, (/nlat,mlon/))   ; degrees of freedom
>  tval = onedtond(rc at tval   , (/nlat,mlon/))   ; t-statistic
> [SNIP]
>
>
> The latter is what Will was suggesting.
>
> On Thu, Jun 22, 2017 at 4:28 AM, Lyndon Mark Olaguera <
> olagueralyndonmark429 at gmail.com> wrote:
>
>> Dear Sir Will and NCL experts,
>>
>> I still get the same error *(fatal:where: dimension sizes  of parameter
>> 0 and parameter 2 do not match)*. I am not sure if I am implementing
>> this correctly.
>>
>> I changed this part in my code based on the suggestion suggestion:
>>
>> ;****************************************************
>> ;Test for significance
>> ;****************************************************
>>  df = rc at nptxy-2                 ; degrees of freedom
>>  tval = rc at tval                  ; t-statistic
>>  b = tval                        ; b must be same size as tval (and df)
>>  b = 0.5
>>  prob = betainc(df/(df+tval^2),df/2.0,b)
>>  printVarSummary(prob);dim=1 size=64800
>>
>> * prob_test =onedtond(prob,(/2,64800/))*
>> * printVarSummary(prob_test)*
>>
>>  rc95 = where(prob_test.lt.0.95,rc at _FillValue,rc)
>>  copy_VarCoords(rc, rc95)
>>  printVarSummary(rc95)
>>
>>
>> I'll appreciate any suggestion on how I can do this correctly.
>>
>> Sincerely,
>>
>> *Lyndz*
>>
>> On Thu, Jun 22, 2017 at 3:34 PM, Will Hobbs <will.hobbs at utas.edu.au>
>> wrote:
>>
>>> Change the dimensions of ‘prob’ so that it matches the dimensions of rc.
>>>
>>>
>>>
>>> You can do this using the onedtond() function; you’ll find a complete
>>> description of this function on the NCL web page.
>>>
>>>
>>>
>>> Will
>>>
>>>
>>>
>>> *From: *<ncl-talk-bounces at ucar.edu> on behalf of Lyndon Mark Olaguera <
>>> olagueralyndonmark429 at gmail.com>
>>> *Date: *Thursday, 22 June 2017 at 4:28 PM
>>> *To: *"ncl-talk at ucar.edu" <ncl-talk at ucar.edu>
>>> *Subject: *[ncl-talk] Question about regressing a time series and a
>>> gridded data using regCoef_n
>>>
>>>
>>>
>>> Dear NCL experts,
>>>
>>>
>>>
>>> I am trying to calculate the linear regression coefficient between a
>>> single time series and a gridded sst using the regCoef_n in ncl. I am using
>>> ncl v.6.4.
>>>
>>>
>>>
>>> Attached is my script and the text file that I am regressing. Here is
>>> the link to the sst data:
>>>
>>>
>>>
>>> [Link to the data]
>>>
>>> https://drive.google.com/file/d/0B9faET7Bc2o8TjNSWnd1ODREcVU
>>> /view?usp=sharing
>>>
>>>
>>>
>>> What I wanted to do:
>>>
>>>
>>>
>>> 1. I want to extract and plot significant coefficients (above 95%
>>> confidence level), so I have this line in the script:
>>>
>>>
>>>
>>>  df = rc at nptxy-2 ; degrees of freedom
>>>
>>>  tval = rc at tval ; t-statistic
>>>
>>>  b = tval ; b must be same size as tval (and df)
>>>
>>>  b = 0.5
>>>
>>> * prob = betainc(df/(df+tval^2),df/2.0,b) *
>>>
>>> * rc95 = where(prob.lt.0.95,rc at _FillValue,rc)*
>>>
>>>
>>>
>>> I encountered an error after this:
>>>
>>>
>>>
>>> *fatal:where: condition variable (parameter 0) dimension mismatch with
>>> parameter 2*
>>>
>>>
>>>
>>> I understand that the *"prob"* variable has only one dimension and the "
>>> *rc"* has 2 dimensions (lat & lon).
>>>
>>>
>>>
>>> Is it possible to assign coordinates to "prob" similar to "rc" other
>>> than the copyVarCoords?
>>>
>>>
>>>
>>>
>>>
>>> I'll appreciate any suggestions on how to do this correctly.
>>>
>>>
>>>
>>>
>>>
>>> Sincerely,
>>>
>>>
>>>
>>> Lyndz
>>>
>>>
>>>
>>> University of Tasmania Electronic Communications Policy (December,
>>> 2014).
>>> This email is confidential, and is for the intended recipient only.
>>> Access, disclosure, copying, distribution, or reliance on any of it by
>>> anyone outside the intended recipient organisation is prohibited and may be
>>> a criminal offence. Please delete if obtained in error and email
>>> confirmation to the sender. The views expressed in this email are not
>>> necessarily the views of the University of Tasmania, unless clearly
>>> intended otherwise.
>>>
>>> _______________________________________________
>>> ncl-talk mailing list
>>> ncl-talk at ucar.edu
>>> List instructions, subscriber options, unsubscribe:
>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>>
>>>
>>
>> _______________________________________________
>> ncl-talk mailing list
>> ncl-talk at ucar.edu
>> List instructions, subscriber options, unsubscribe:
>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20170623/5b9da8b2/attachment.html