Re: Need a help for a logic of escorc

Thanks, but that's not what I meant. I can handle the above case easily..,

My problem is : I just wonder how to *create a series* where Y shifts
according to the lag in each year dataset. then hold them together to be
one series. X is kept fixed 90 days (only in JFM) for each year (i.e only Y
is moving back and forward, meaning that Y time series start on 17 December
and 15 January in analysis for each year)

The illustration is as follow:

lag 0 :

X =
1Jan1980---------------31mar1980||1Jan1981--------------31mar1981--------------

Y =
1Jan1980---------------31mar1980||1Jan1981--------------31mar1981--------------

lag +1 :

X =
1Jan1980---------------31mar1980||1Jan1981--------------31mar1981--------------

Y =
2Jan1980---------------1Apr1980||2Jan1981------------------1Apr1981--------------

lag -1 :

X =
1Jan1980---------------31mar1980||1Jan1981--------------31mar1981---------------

Y =
31Dec1979---------------30mar1980||31Dec1980-------------30mar1980--------------

surely each segment for one year is 90 days, so if have 20 years data, the
total length of series will be 20 x 90 =1800 time.

FOR X is quite EASY to handle, but FOR Y is quite hard to hande by shift
the date accordingly for each year and combine them. That's actually my
problem.

Thank you.
Erika

On Wed, Aug 14, 2013 at 5:33 PM, Dennis Shea <shea@ucar.edu> wrote:

> Please do not send me direct emails. I have my own job to do. THX.
>
> ===
>
> Alessandra's response was "appropriate". However, another approach
> may be the following where *you* 'manually' do the lag operation.
>
> Note: escorc requires that the number of time steps be the same.
>
> Something like the following.
>
> X = (time,lat,lon)
> Y = (time,lat,lon)
>
> undef("erika_lag")
> function erika_lag(X[*][*][*], Y[*][*][*], mxlag[1]:integer)
> begin
> x = X(lat|:,lon|:,time|:)
> y = Y(lat|:,lon|:,time|:)
>
> dimx = dimsizes(x)
> nlat = dimx(0)
> mlon = dimx(1)
> ntim = dimx(2)
> ntim1 = ntim-1 ; convenience
>
> mxlag = 15
> r = new ( (/2*mxlag+1,nlat,mlon/), typeof(x), getFillValue(x))
>
> r(mxlag+1,:,:) = escorc(x,y) ; correlation at lag 0
>
> do m=1,mxlag ; x lead y
> myStrt = m
> myLast = ntim1
> r(mxlag+1+m,:,:) = escorc(x(:,:,0:ntim1-m),y(:,:,**myStrt:myLast))
> end do
>
> do m=1,mxlag ; y lead x
> myStrt = ?
> myLast = ?
> r(m-mxlag-1,:,:) = escorc(y(:,:,myStrt:myLast)), x(:,:,0:ntim1-m))
> end do
>
> r!0 = "lag"
> r&lag = ispan(-mxlag,mxlag,1) ; -15 to 15
> copy_VarCoords(x(:,:,0), r(0,:,:))
> printVarSummary(r)
>
> return(r)
> end
>
>
>
>
>
> On 8/14/13 6:01 AM, Erika Folova wrote:
>
>> Dear NCL Team,
>>
>> It was posted a week ago but haven’t got any appropriate response yet -(
>> I am begging a suggestion for this tricky problem.
>>
>> Actually the question is how can I* control the date* for Y. For X is OK
>> to
>> keep it *fixed 90 days* for each year from 1 January to 30 March. But for
>> Y
>>
>> is quite hard for me to understand how can I shift it by -15 to 15 days
>> relative to X, meaning that Y time series start on 17 December and 15
>> January for the first and the last analysis.
>>
>> Here is the illustration of the setting of time series :
>>
>> lag 0: X: 1 Jan |----------------------| 31 Mar 1
>> Jan|----------------------|31 Mar .....
>> Y: 1 Jan |----------------------| 31 Mar 1
>> Jan|----------------------|31 Mar .....
>>
>> lag -1 X: 1 Jan |----------------------| 31 Mar 1
>> Jan|----------------------|31 Mar ....
>> Y: 31 Dec |----------------------| 30 Mar 31
>> Dec|----------------------|30 Mar ....
>>
>> lag +1 X: 1 Jan |----------------------| 31 Mar 1
>> Jan|----------------------|31 Mar ....
>> 2 Jan |----------------------| 1 Apr 2
>> Jan|----------------------| 1 Apr ....
>>
>>
>> Thank you so much for your kind suggestion,
>> It means a lot for me to figure out this matter.
>> Thank you
>>
>> Best,
>> Erika
>>
>>
>>
>>
>>
>>
>> On Fri, Aug 9, 2013 at 8:54 PM, Erika Folova <e.folova@gmail.com> wrote:
>>
>>> I meant as I described earlier, I want to keep 90-day time period for X
>>> "fixed" from1 January to 31 March while the time series of the Y is
>>>
>> shifted
>>
>>> by -15 to 15 days (nlag= +/- 15 ). That meansthe Y time series start on
>>> 17
>>> December and 15 January for the first and the last analysis. I still do
>>>
>> not
>>
>>> have an idea how to make it easily. Any help will be really appreciated.
>>>
>>> Thank you,..
>>>
>>>
>>>
>>>
>>> On Fri, Aug 9, 2013 at 2:42 AM, Alessandra Giannini
>>> <alesall@iri.columbia.edu> wrote:
>>>
>>>>
>>>>
>>>> Hi Erika,
>>>>
>>>> NCL has a correlation function that allows for lags:
>>>>
>>>> <http://www.ncl.ucar.edu/**Document/Functions/Built-in/**esccr.shtml<http://www.ncl.ucar.edu/Document/Functions/Built-in/esccr.shtml>
>>>> >
>>>>
>>>> cheers, alessandra
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Aug 8, 2013, at 1:38 PM, Erika Folova <e.folova@gmail.com> wrote:
>>>>
>>>> So.. no suggestion for my case :-(
>>>>
>>>>
>>>> On Sat, Jul 27, 2013 at 5:58 AM, Erika Folova <e.folova@gmail.com>
>>>> wrote:
>>>>
>>>>>
>>>>> Hallo NCL,
>>>>>
>>>>> Lately I am quite confused to set my code to do a cross correlation,
>>>>> The role is as follows: I have daily datasets lets say X and Y
>>>>> extending for 5 years. The role is, I want to keep 90-day time period
>>>>>
>>>> for
>>
>>> X "fixed" from
>>>>> 1 January to 31 March while the time series of the Y is shifted by -15
>>>>>
>>>> to
>>
>>> 15 days (nlag= +/- 15 ). That means
>>>>> the Y time series start on 17 December and 15 January for the first and
>>>>> the last analysis.
>>>>>
>>>>> What I think is something like this:
>>>>>
>>>>> mxlag = 15
>>>>> lag = ispan(-mxlag,mxlag,1)
>>>>>
>>>>> do l = -mxlag,mxlag
>>>>> CC (l+mxlag,:) = escorc( X(time), Y(time|time+l))
>>>>> end do
>>>>>
>>>>> But I guess this is a wrong approach because I just want to concern on
>>>>> JFM fixed on X but shifted for Y as I mentioned previously,
>>>>> Any help would be really appreciated.
>>>>>
>>>>> Thank you
>>>>>
>>>>> Erika,
>>>>>
>>>>
>>>>
>>>> ______________________________**_________________
>>>> ncl-talk mailing list
>>>> List instructions, subscriber options, unsubscribe:
>>>> http://mailman.ucar.edu/**mailman/listinfo/ncl-talk<http://mailman.ucar.edu/mailman/listinfo/ncl-talk>
>>>>
>>>>
>>>>
>>>
>>

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Wed Aug 14 10:25:59 2013