[gradsusr] correlation difference between GrADS and NCL

[Andrew Friedman]

This should allow you to isolate whether the issue is GrADS or the implementation.
Andrew

On Jun 13, 2014, at 1:09 AM, Andrew Friedman <andfried at berkeley.edu> wrote:

> Hi Rui,
> 
> As a more simple check than another program, can you compare your GrADS results with a simple correlation coefficient calculated using the standard formula with a handful of data points?
> https://en.wikipedia.org/wiki/Pearson_product-moment_correlation_coefficient 
> 
> Andrew
> 
> On Jun 12, 2014, at 3:27 PM, Cheng, Rui <r5cheng at ucsd.edu> wrote:
> 
>> Hi all GrADS users,
>> 
>> I want to get the correlation between dcape and rainrate using GrADS. Meanwhile, NCL is used to test the result.
>> 
>> Assuming dcape is one time step prior to rainrate, so we can get the correlation (0.0266) using GrADS through,
>> "set t 2 77; define dcapeave2=dcapeave(t-1); set t 1; d tcorr(dcapeave2,rainrateconvm,t=2,t=77)."
>> 
>> Also using NCL, we can get the correlation (-0.18815) through,
>> "ccr2=esccr(dcapeave(1:76,0,0,0),rainrateconvm(1:76,0,0,0),1)."
>> 
>> I notice that the difference is small if there is no time lag between dcape and rainrate. Then we get the correlation "-0.1139" using GrADS and "-0.1101" using NCL.
>> 
>> Why the difference is so big when the time lag is present between two time series? 
>> Would you like to help me?
>> 
>> Many thanks,
>> 
>> Rui
>> 
>> _______________________________________________
>> gradsusr mailing list
>> gradsusr at gradsusr.org
>> http://gradsusr.org/mailman/listinfo/gradsusr
>