Re: T-Test function in NCL

From: Dennis Shea <shea_at_nyahnyahspammersnyahnyah>
Date: Tue Nov 12 2013 - 19:33:55 MST

A single precision (NCL uses double precision) of the code is at:
    http://www.cgd.ucar.edu/~shea/ttest_ncl.f

It returns the same answers as (say) Numerical Recipes.

http://www.cgd.ucar.edu/~shea/ttest_ncl.f

On 11/12/13 6:16 PM, Heidi Stauffer wrote:
> Hello,
>
> I attempted to use the ncl t-test function to test the significance of
> results from a sensitivity test using a regional climate model. Prior to
> this, I extracted output from two specific locations and conducted a
> t-test using R and the results were significant for both locations. I
> had noticed that the variances appeared quite different for the two
> model scenarios so I used the Welch's t-test in R (which, unlike the
> Students t-test assumes that the variances of the two samples are not
> the same).
>
> The output is in the form of seasonal averages over 25 years. The files
> for the two model scenarios have been uploaded to the ftp site and are
> entitled: SRF.wet8510.nc and SRF.wet8510N.nc
>
> However, when I attempted to use the ncl function, the resulting plot
> (and raw probability results) showed very few areas of significant
> difference over the domain, and none of those areas overlapped with the
> two locations I independently tested. I've attached my script and a
> sample plot of the raw probability results.
>
> I am wondering if someone can tell me how the ncl function actually
> works (if I could possibly see the original code), especially how iflag
> works and how the ttest function calculates variances. Also, if the
> option for different variances is working the same way as the Welch's
> t-test. If the option isn't working, I was hoping that someone has
> written an alternative script which incorporates the Welch's t-test or
> similarly accounts for different variances.
>
> Thank you.
>
> Regards,
> Heidi
>
>
>
> _______________________________________________
> 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 Tue Nov 12 19:34:01 2013

This archive was generated by hypermail 2.1.8 : Fri Nov 22 2013 - 09:36:32 MST