Re: normalization in cospc from specxy_anal [SEC=UNCLASSIFIED]

The cospectrum returned by specxy as an attribute ('cospc')
is a spectral density.

If you want the area under the cospc 'curve' to be equal
to the covariance, you can use
   http://www.ncl.ucar.edu/Document/Functions/Built-in/escovc.shtml
to calculate the (sample) covariance of the detrended
series (eg, 'COV'); then normalize such that

  COV= SUM(cospc(1:n-2)*df) + (cospc(0) + cospc(n-1))*df*0.5

On 11/25/11 4:10 PM, Eun-Pa Lim wrote:
> Thanks for your reply, Dennis!
>
> I meant covariance to be x*y where x and y are detrended time series.
> If it's not the way to calculate cospectra power density,
> is there any other function/procedure or ncl script I can use?
>
> Cheers
>
> Eunpa
> __
>> If I want to get cospectra power density, can I calculate it from the
>> same procedure as spcx?
>> e.g. cospectrum*covariance/(total area under the curve)
>
> Not sure ... what is "covariance" ?
>
>> Thanks!!
>> Eunpa
>>
>> __________________________________________________
>>
>> *Eun-Pa Lim*; CAWCR
>> GPO Box 1289K, Melbourne 3001, VIC, AUSTRALIA
>> Phone +61 3 9669 4636; Fax +61 3 9669 4660 ; http://www.bom.gov.au/bmrc/
>> <http://www.bom.gov.au/bmrc/>******
>>
>>
>>
>> _______________________________________________
>> ncl-talk mailing list
>> List instructions, subscriber options, unsubscribe:
>> 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 Mon Nov 28 09:32:01 2011