[ncl-talk] Query regarding ttest and pdfx function

[Dennis Shea]

Good questions. I'm not a statistician so I'm not sure if I have good
answers. I think it is best to consult a statistician. A few random
thoughts:

[1.1] ave1, ave2: averages of the 2002 JJAS season [122 days]
[1.2] var1, var2: variances " ...."
[1.3] s1, s2: Typically, in* mid-latitudes*, the time between independent
estimates [aka: the characteristic time: T0] for 'continuous' quantities
like temperature and (say) sea-level-pressure is 4-to-7 days. These are
derived by assuming an autoregressive process. Regions of frequent
cyclogenesis have the smaller values. These estimates could be derived for
every grid-point. However, for simplicity and convenience and  being
somewhat conservative [T0=6]:  s1=N/T0=122/6 = 20 is a good estimate for
*each* seasonal time series.
====
PDF: If you have a region: latS:latN,  lonE:lonE

I guess it depends on what you are looking for.

PDF_all  = *pdfx*
<https://www.ncl.ucar.edu/Document/Functions/Contributed/pdfx.shtml>(bias(:,{latS:latN},{lonW:lowE}),
nbin, opt)   ; over all time and region
PDF_specific_time = *pdfx*
<https://www.ncl.ucar.edu/Document/Functions/Contributed/pdfx.shtml>(bias(nt,{latS:latN},{lonW:lowE}),
nbin, opt)   ; specific time [nt] for a region
PDF_specific_grid_point: *pdfx*
<https://www.ncl.ucar.edu/Document/Functions/Contributed/pdfx.shtml>(bias(:,{LAT},{LON}),
nbin, opt)  ; all time for a specific location

===
Good luck

On Sun, Feb 23, 2020 at 6:59 AM Dr. Suman Maity via ncl-talk <
ncl-talk at ucar.edu> wrote:

> Dear all
> I am new for ncl and recently started using it. However, I have the
> following queries:
> 1. I have temperature data from the model (time,lat,lon) and observation
> (time,lat,lon) with time=122 days (JJAS), lon=692 grid-points and lat=466
> grid-points. I want to plot JJAS bias (model-observation) over-layed with
> 95% significant bias gridpoints as dots.  I took the help of "ttest"
> function in ncl. My script is attached herewith. I would like to request
> you to check the script and modify if needed. However, general syntax of
> ttest is like the following:
> prob = ttest(ave1,var1,s1,ave2,var2,s2,iflag,tval_opt)
> What is this "s1" and "s2" ? it is mentioned the they gives no. of
> statistically independent observation. Am I need to calculate that using
> "equiv_sample_size" function? or any constant value will be fine? if yes
> what will be the approximate values?
>
> 2. I want to generate probability density function of temperature bias
> over a region. I have bias data(time,lat,lon). My query is:
> 1) pdfx function is generally applied on which dimension? I mean time or
> space?
> 2) Is the rand_normal is required in case of my data?
> 3) In my case bias (model-obs) is a time(122 days) X lon X lat dimension
> data. How pdfx function will be applied on that? Will it directly go to the
> pdfx function like:
> ap = pdfx(bias,0,True)
> or
> I need to average the bias w.r.t time and then apply  ap =
> pdfx(bias,0,True)
> Any sort of suggestion will be highly appreciated.
> Best
> Suman
> ++++++++++++++++++++++++++++++++++++++++++
> Dr. Suman Maity
> Research Associate
> School of Atmospheric Science
> Sun Yat-sen University, Zhuhai campus,Tangjiawan Town
> Zhuhai, Guangdong, Postcode-519082,China.
> Phone:+86-13543861045 (China), +91-9732636778 (India)
> skype: reach2suman
> https://www.researchgate.net/profile/Suman_Maity
> ++++++++++++++++++++++++++++++++++++++++++
> _______________________________________________
> ncl-talk mailing list
> ncl-talk at ucar.edu
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20200224/5439df1d/attachment.html>