Re: [ferret_users] Different correlation values obtained when order of variable changed

["Ansley C. Manke" <ansley.b.manke@xxxxxxxx>]

Hi,

Yes, the results should be the same.  List out your P and Q variables.  Are they actually the same in each case, with P and Q reversed?  I know they are the same in your commands below, but maybe in your script there's a difference that's hard to see.

-Ansley

On 12/2/2016 1:55 PM, afwande juliet wrote:

Dear ALL
I am using regressxy to calculate correlation coefficients over areal average.
when i change the order of p and q , the values significantly varies, yet I thought the order of the variables doesn’t matter

for example when I do like this ...

let p =pr[d=3,x=30E:42E@ave, y=0:5N@ave,l=1:3]
let q =pr[d=1,x=30E:42E@ave, y=0:5N@ave,l=1:3]
go variance

list correl =0.5

but when I reverse the order say ...

let q =pr[d=3,x=30E:42E@ave, y=0:5N@ave,l=1:3]
let p =pr[d=1,x=30E:42E@ave, y=0:5N@ave,l=1:3]
go variance
list correl =0.1