Re: Appropriate significant test (?)

I think your question is: "What combination of dates could be used
to a priori create a suite of dates that would yield
significant differences?"

This is not the way 'statistical tests' are done.
Usually the dates are specified a priori on the basis of some
physical situation. For climate, think El Nino & La Nina.
Then the result is tested at some confidence level specified
a priori. I think you should do the same.

If you are randomly searching for some dates the only
way I can think of is to randomly select the central dates
and place one in one category; the next in another;
perform calculations; if significant at 95% level; save the
dates.

See
http://www.ncl.ucar.edu/Document/Functions/Contributed/generate_unique_indices.shtml

to generate the indices.

===
Good Luck

On 1/15/14, 7:32 AM, Ahmad Farsyud wrote:
> Dear NCL,
>
> I am normally using "ttest" function in NCL for calculating a significant
> test for my composite. However, in one point I'm not really sure whether
> what I've done is entirely correct or not.
>
> Let's say I have two years daily data 2006-2007 of U [time, lat, lon], in
> total I have 730 days for my data. Within this period, I have 25 selected
> central dates varied within this period. My question is, how could I
> perform a significant test so that this compositing event are statistically
> significant at 95% level ?
>
> Please share your expertise.
> Thank you
>
> -- AF
>
>
>
> _______________________________________________
> 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 Wed Jan 22 14:27:32 2014