# How to balance two Forex crosses correctly to do a linear regression?

I have two cross and an account in EUR:

1. EUR/USD
2. GBP/USD

I would like to do a balanced linear regression using R. With "balanced" I mean that I would like to normalize it by calculating the difference in pips from the previous day and the current, and then multiply by the respective (middle) pip value.

I do a simple example:

The series are:

EUR/USD 1.3000 1.3050 1.3060

GBP/USD 1.6000 1.6050 1.6060

The differences are:

EUR/USD

1. 50 pips (1.3000 - 1.3050)
2. 10 pips (1.3050 - 1.3060)

GBP/USD

1. 50 pips (1.6000 - 1.6050)
2. 10 pips (1.6050 - 1.6060)

Now, calculate (the values are inventend) the middle pip values:

(remember that account currency in EUR)

EUR/USD

1. 13 € pip value (at middle pip: 1.3025)
2. 13.5 € pip value (at middle pip: 1.3055)

GBP/USD

1. 9 € pip value (at middle pip: 1.6025)
2. 8.5 € pip value (at middle pip: 1.6055)

Now the normalized series that I studied should be:

EUR/USD

• 50(pip difference) * 13€ = 650
• 10(pip difference) * 13.5€ = 135

GBP/USD

• 50(pip difference) * 9€ = 450
• 10(pip difference) * 8.5€ = 85

The normalized series are:

EUR/ USD: 650, 135

GBP/USD: 450, 85

What do you think of this type of procedure? (then I will use those series in the linear regression)

-
Hopefully I'm beginning to understand but I wonder: could you also use a return series? – Bob Jansen Dec 3 '11 at 13:41
@Bootvis If I use the returns series I think that the coefficients of the linear regression will be "wrong", I mean, If I use returns instead of the procedure above I could get different returns BUT with the same "move" (the same number of pips)...so when I will look at the coefficients of the linear regression, I can't use the coefficient AS multiplier for lots. No? – Dail Dec 3 '11 at 21:57