Re: RMS and named subscripting

Dear Dennis.
Thank you for your advice for debugging my error and is graciously received
Your last suggestion is very helpful.

But, please help me understand this because I did review your documentation
before submitting my question.

If I did not want to use the onedtond function, per your last suggestion,
how would I calculate the rmsd in same manner that I calculate the
unweighted sample cross-correlation here:

 ; Array of unweighted spatial Correlation Coefficients
fin925_R_arr = new((/ntim/),"float")
 ; Array of unweighted spatial RMS Difference
rmsdTime925=new((/ntim/),"float")

do nt = 0, ntim-1
 ; Calculate unweighted spatial Correlation Coefficients
 fin925_R_arr(nt) = escorc(ndtooned(V925(nt,:,:)),ndtooned(FNLv925(nt,:,:)))
 ; Calculate RMS
 ; rmsdTime925(nt) =
End do

***Since both functions expect a "numeric" float or double,.....
escorc
Computes sample cross-correlations at lag 0 only.
Prototype

    function escorc (
        x : numeric,
        y : numeric
    )

    return_val : numeric
AND
dim_rmsd_Wrap
Computes the root-mean-square-difference between two variables' rightmost
dimension at all other dimensions.
Prototype

    function dim_rmsd_Wrap (
        x : numeric,
        y : numeric
    )

    return_val : float or double

..... I thought I could calculate the RMS the same way as the correlation.
So I posted my question as such (but using the named dimensions).

Sincerely,
Erik

On 2/7/08 9:01 AM, "Dennis Shea" <shea_at_ucar.edu> wrote:

> [1] Please examine the error message [as previously suggested by
> Mary in another response]
> [2] always read the function documentation.
>
> Prior to the do loop *you* defined
> rmsdTime925=new((/nvtimes/),"float") ; named RMS array
> This is one dimensional.
>
> The offending statement:
> rmsdTime925(nt) =
> dim_rmsd_Wrap(FNLv925(lat|:,lon|:,time|(nt)),V925(lat|:,lon|:,time|(nt)))
>
> The error message:
> "fatal:Dimension sizes on right hand side of assignment do not match
> dimension sizes of left hand side"
>
> To debug you should look at the error message and say to yourself
> "Ummm, I have specified that the left hand side is size 1 [a scalar]
> rmsdTime925(nt)
> The error message says that the right hand size does not match
> this expectation. I should read the documentation which has several
> example illustrating returned dimension sizes or I can debug myself
>
> Just before the offending statement I'll add
>
> debug =
> dim_rmsd_Wrap(FNLv925(lat|:,lon|:,time|(nt)),V925(lat|:,lon|:,time|(nt)))
> printVarSummary(debug)
>
> You will see that "debug" is not a scalar as you expect.
>
> Finally,
>
> dim_rmsd_Wrap(FNLv925(lat|:,lon|:,time|(nt)),V925(lat|:,lon|:,time|(nt)))
>
> the last argument is a scalar [ time(nt) ]. The dim_rmsd expects an vector
> as the rhs so, in essense, your statement reduces to
>
> dim_rmsd_Wrap(FNLv925(lat|:,lon|:),V925(lat|:,lon|:))
>
> ====
> If you want an unweighted spatial rmsd at each time step
> and nlat and mlon are the size of the lat and lon dimensions,
> and you want to keep it in the loop
>
> dum1 = onedtond( ndtooned(FNLv925(nt,:,:)), (/1, nlat*mlon/))
> dum2 = onedtond( ndtooned(v925(nt,:,:)) , (/1, nlat*mlon/))
>
> rmsdTime925(nt) = dim_rmsd_Wrap(dum1,dum2)
>
>
> Erik Noble wrote:
>> I am following onto this question:
>> I used named subscripting in my code and it worked.
>> I tried using the dim dim_rmsd_Wrap function and I received only this error:
>>
>> Copyright (C) 1995-2007 - All Rights Reserved
>> University Corporation for Atmospheric Research
>> NCAR Command Language Version 5.0.0
>> The use of this software is governed by a License Agreement.
>> See http://www.ncl.ucar.edu/ for more details.
>>
>> fatal:Dimension sizes on right hand side of assignment do not match
>> dimension sizes of left hand side
>> fatal:Execute: Error occurred at or near line 86 in file test_script2.ncl
>>
>>
>> In the code where the error occurs is
>> ; Rename dimensions for clarity's sake
>> FNLv925!0 = "time"
>> FNLv925!1 = "lat"
>> FNLv925!2 = "lon"
>> FNLv700!0 = "time"
>> FNLv700!1 = "lat"
>> FNLv700!2 = "lon"
>> V925!0 = "time"
>> V925!1 = "lat"
>> V925!2 = "lon"
>> V700!0 = "time"
>> V700!1 = "lat"
>> V700!2 = "lon"
>> rmsdTime925=new((/nvtimes/),"float") ; named RMS array
>>
>> ; Correlation
>> do nt = 0, nvtimes-1
>> fin925_R_arr(nt) = escorc(ndtooned(V925(nt,:,:)),ndtooned(FNLv925(nt,:,:)))
>> fin700_R_arr(nt) = escorc(ndtooned(V700(nt,:,:)),ndtooned(FNLv700(nt,:,:)))
>> ; RMS Error
>> rmsdTime925(nt) = dim_rmsd_Wrap(
>> FNLv925(lat|:,lon|:,time|(nt)),V925(lat|:,lon|:,time|(nt)))
>> end do
>>
>> How does my rms function differ that far from the example (at
>> http://www.ncl.ucar.edu/Document/Functions/Contributed/dim_rmsd_Wrap.shtml)
>> to where I am getting a dimension size error?
>>
>>
>> On 2/6/08 10:41 PM, "Erik Noble" <enoble_at_giss.nasa.gov> wrote:
>>
>>
>>> Hi.
>>> I want to compute the temporal root-mean-square-difference at each
>>> latitude/longitude grid point and I know that I have to follow the NCL's
>>> named subscripting to reorder the input array such that "time" is the
>>> rightmost dimension, such that
>>> rmsdTime = dim_rmsd( x(lat|:,lon|:,time|:), y(lat|:,lon|:,time|:) )
>>>
>>> But my two variables are:
>>>
>>> Variable: V925
>>> Type: float
>>> Total Size: 3636240 bytes
>>> 909060 values
>>> Number of Dimensions: 3
>>> Dimensions and sizes: [60] x [109] x [139]
>>> Coordinates:
>>> Number Of Attributes: 4
>>> PlotLevelID : 925 hPa
>>> units : m/s
>>> description : v Velocity
>>> _FillValue : -999999
>>>
>>>
>>> Variable: FNLv925
>>> Type: float
>>> Total Size: 3636240 bytes
>>> 909060 values
>>> Number of Dimensions: 3
>>> Dimensions and sizes: [60] x [south_north_stag | 109] x [west_east | 139]
>>> Coordinates:
>>> Number Of Attributes: 6
>>> stagger : V
>>> description : V
>>> units : m s-1
>>> MemoryOrder : XYZ
>>> FieldType : 104
>>> _FillValue : -999
>>>
>>>
>>> Since, FNLv925's dimensions have names and V925's dimensions do not, in
>>> order to root-mean-square-difference at each latitude/longitude grid point,
>>> do I need to add extra dimension names to V925?
>>> Thank you.
>>> -Erik
>>>
>>>
>>> _______________________________________________
>>> ncl-talk mailing list
>>> ncl-talk_at_ucar.edu
>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>>
>>>
>>
>>
>> _______________________________________________
>> ncl-talk mailing list
>> ncl-talk_at_ucar.edu
>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>
>
>
>

_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Thu Feb 07 2008 - 13:57:45 MST