## Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w. Y?

Timeseries X1 and X2 are both highly correlated with Y (Cov>90%), why the difference btw X1 and X2 are still highly

correlated with Y(COV~=90%)?

How to treat this to make a better prediciion?!

7 REPLIES 7

## Re: Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w

What does that imply?! Or just typical?!

## Re: Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w

Please explain further the part in red ... difference of what?

“...why the difference btw X1 and X2 are still highly correlated with Y(COV~=90%)?”

--
Paige Miller

## Re: Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w

New variable =X1-X2 still correlated with Y, why? Typical? Or abnormal?

## Re: Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w

Can you show us the correlation matrix of X1, X2, (X1-X2) and Y

--
Paige Miller

## Re: Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w

Spoiler
Y=h300 x1=ordvol_dif_gap4 x2=ordvol_dif_gap1 x1-x2=ordvol_dif_g14
Pearson Correlation Coefficients, N = 22999
Prob > |r| under H0: Rho=0
hs300 ordvol_dif_g14 ordvol_dif_gap4 ordvol_dif_gap1
hs300
 1.00000
 0.95534 <.0001
 0.94302 <.0001
 0.93251 <.0001
ordvol_dif_g14
 0.95534 <.0001
 1.00000
 0.98923 <.0001
 0.97906 <.0001
ordvol_dif_gap4
 0.94302 <.0001
 0.98923 <.0001
 1.00000
 0.99831 <.0001
ordvol_dif_gap1
 0.93251 <.0001
 0.97906 <.0001
 0.99831 <.0001
 1

## Re: Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w

I don't think there's anything unusual here, did you try plotting the data?

--
Paige Miller  Ksharp
Super User

## Re: Confusion, why difference between two X Vars(highly correlated with Y) still highly correlated w

That is normal. Make the second order difference and Check Corr ,if it still high , and make the third difference ,until it is not correlation.
Discussion stats
• 7 replies
• 287 views
• 0 likes
• 3 in conversation