>dear ncl friends:
> i am just a beginner for using ncl.i have some problem when i use ncl for
spectrum analysis.
> 1.when we use the function specx_anal,which number could i choose for its
third dimension "jave" ?i meant that i don not know what should be depentent on
for the choice,or just only satifying "at least 3 or odd"
jave: The choice of jave affects the bandwidth, the character of
the spectrum and the number of the degrees of fredom.
smaller jave yields higher resolution in the frequency domain
(smaller bandwidth). The spectrum will look more jagged
and the number of degrees of freedom will be less.
larger jave yields broader resolution in the frequency domain
(larger bandwidth). The spectrum will look more smoother
and the number of degrees of freedom will be larger.
Spectral analysis is part-art, part-science. The choice
of jave is a funtion of what you are looking for.
> 2,when i use the function specx-ci for red noise significent test,why is the
lower bound should be drawn in its output ?
Because it is a two-sided test.
> in my eyes,we often use the upper bound but not the lower,could you tell me
the reason for the lower bound line?
Because variance less than that expected by a 1st-order (ie, red-noise)
Markon model is as significant as more variance expected.
> 3,another problem is that if i want to write freqency in the below x axis
,meanwhile i also want to write the relative period in the top x axis,what can i
do in ncl ?(the frequecy and the period are corresponding)
Are u asking about:
period
__________________________________
| |
| |
S | |
| |
| |
-----------------------------------
frequency
[1] You can copy and modify the spectral analysis code.
[2] In the next version of NCL, [a032], a new gsn_csm
function gsn_csm_x2y (wks, freq, period, spec, r1, r2)
good luck
_______________________________________________
ncl-talk mailing list
ncl-talk@ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
This archive was generated by hypermail 2b29 : Fri Feb 27 2004 - 09:12:21 MST