[ncl-talk] plot lead/lag correlation with statistical significance

[Dennis Shea]

Yes. No significance would be my assessment.

I speculate there are refereed papers on this topic.



On Mon, Nov 26, 2018 at 12:48 AM Sujata Mandke <amin at tropmet.res.in> wrote:

> Respected Dennis Shea sir,
>
>
> I have included all your suggestions in the ncl script (modified ncl
> script is attached).
>
>
> The correlation plot using this ncl script is also attached.
>
>
> For statistical significance - The estimate of number of independent
> values
>
> using "equiv_sample_size
> <http://www.ncl.ucar.edu/Document/Functions/Built-in/equiv_sample_size.shtml>"
> for snow over Eurasia =64 and for SST over
>
> East Pacific =7 . Minimum of the two, that is 7 is used for statistical
> significance
>
> testing using "rtest". Probability values representing statistical
> significance of
>
> correlation (for maximum lag=12) are high (ascii file for the same is
>
> attached for total lag=12*2+1) and therefore none is
>
> significant (not even near significance).
>
>
> Does this indicate that none of the correlation is statistically
>
> significant or there is a problem in the way it is estimated?
>
> With best regards
>
> Dr. Sujata Mandke
>
> scientist, IITM,PUNE, INDIA
>
>
> ------------------------------
> *From: *"Dennis Shea" <shea at ucar.edu>
> *To: *"S.Amin" <amin at tropmet.res.in>
> *Cc: *"ncl-talk" <ncl-talk at ucar.edu>
> *Sent: *Wednesday, November 21, 2018 3:34:30 AM
> *Subject: *Re: [ncl-talk] plot lead/lag correlation with statistical
> significance
>
> My opinion:
>
> [1] You should remove the climatological annual cycle from each dataset.
>
> *http://www.ncl.ucar.edu/Document/Functions/Contributed/clmMonTLL.shtml*
> <http://www.ncl.ucar.edu/Document/Functions/Contributed/clmMonTLL.shtml>
>
> *http://www.ncl.ucar.edu/Document/Functions/Contributed/calcMonAnomTLL.shtml*
> <http://www.ncl.ucar.edu/Document/Functions/Contributed/calcMonAnomTLL.shtml>
>
>
> [2] Typically, the 'swe' file contains a _FillValue (-999.0) when there is
> no snow cover. I suggest setting this  to 0.0.
>
>    sn= in1->swe(iStrt:iLast,{50:70},{20:140})  ;only specific region
> averaged over total Eurasia
>    sneu r=* wgt_areaave_Wrap*(sn,1.0,1.0,0);;;create area averaged snow
> over Eurasia
>    sneur = *where*(*ismissing*(sneur), 0.0, sneur)
>    *printVarSummary*(sneur)
>    p*rintMinMax(*sneur,0)
>    print("---")
>    ymd    := *cd_calendar*(sneur&time, -2)  ; yyyymmdd
>    print(ymd+"  "+sneur)
>    print("---")
>
> [3] You *know* the climatology. You want to know the feedback of sst/sneur
> anomalies on each other.
> Use the anomalies in the correlation calculations.
>
> [4] There are statistical issues. EG: the number of independent values to
> be used in the testing.
>
> Typically, successive monthly values of SST or SNE are not independent. If
> (say) a January SNEUR is very large, then the February SNE will likely be
> large also. So estimating the number of independent values is an issue.
>
>  *
> http://www.ncl.ucar.edu/Document/Functions/Built-in/equiv_sample_size.shtml
> <http://www.ncl.ucar.edu/Document/Functions/Built-in/equiv_sample_size.shtml>*
>
> Estimate for the SNEUR/SST anomalies separately. Use the smaller value.
>
> [5] I am sure the IITM has staff members who know more about statistical
> testing/inference than ncl-talk. I suggest talking with them.
>
> Good luck
>
> On Tue, Nov 20, 2018 at 6:13 AM Sujata Mandke <amin at tropmet.res.in> wrote:
>
>> Dear NCL community,
>> Greetings!
>>  I have ploted lag/lead correlation (cc) between two area-averaged
>> (Eurasian snow and East Pacific SST) monthly time series as xy plot
>> (line).
>> Further, I want to indicate which part of cc is statistically
>> significant on this line using line markers.
>>
>> I had tested statistical significance of cc using “rtest”.
>>  In the ncl script(attached),  I have plotted  line for
>>  lead/lag correlation. How to mark part of this cc (line),
>> which is statistically significant by using line markers.
>>
>> My questions are:
>> (i) Is my statistical significance testing correct?
>>
>> (ii) how to indicate those points (by marker or any other way),
>>  that are statistically significant on the   lead/lag cc line plot.
>>
>> I had extensively searched NCL-talk archives but
>> did not find solution to my problem. My guess is that,
>>  i had to use “where” function and then gsn_add_polymareker,
>>  but do not know how to implement this in the ncl script.
>> I am using NCL version 6.4.0 on linux machine.
>>
>> Any suggestion would be of great help.
>> Many thanks in advance.
>> With best regards
>> Dr. Sujata Mandke
>> Scientist, Indian Institute of Tropical Meteorology
>> PUNE, INDIA_______________________________________________
>> 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/20181126/004adcb0/attachment.html>