Re: Testing for Autocorrelation

On Sun, 13 Nov 2005, Hui Wan wrote:

> Hello,
>
> I want to test the lag-1 auto-correlation coefficient I've got from the NCL function "esacr" to see if the correlation
> is significant.
>
> For the cross-correlation, one can construct a t-statistic and use function "betainc" to evaluate it. I don't know
> whether the same can be applied to the lag-1 auto-correlation. Has anyone got an idea?
>
> Thanks.
>
> Hui
>
>
>
yes the same approach can be used; [see Numerical Recipies: Linear Correlation]

http://www.ncl.ucar.edu/Document/Functions/Built-in/rtest.shtml

; calculate the significance level
; . [see Numerical Recipies: Linear Correlation]
; . [also: Climatic Change, WMO Tech Note 79, p66
; . says "tval" formula is good for xn>=8]

       dn = n-2.
       tval = r*sqrt(dn/(1.-min(r*r, 0.999)))
       prob = betainc ( dn/(dn+tval**2), 0.5*dn, 0.5 )

_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Sun Nov 13 2005 - 07:51:34 MST