Estimating second-lag PAC coefficients through regression

[ssantic]

Hi!

Full disclosure: this is part of a homework, but only 1/4th of one of 5 excercises. So here goes:

I have two AR(2) time series generated(both are in the attached datafile). Their specifications are:
Xt = 0.2 - 0.6Xt-1 + 0.3Xt-2 + et
Yt = 0.2 + 0.6Xt-1 - 0.3Xt-2 + et

The text of the excercise is this: "using the Eviews 'Correlogram' option, we can easily get estimated of the partial autocorrelation (PAC) functions. Confirm that similar estimates of the partial autocorrelation coefficient on the SECOND lag can be achieved through a regression approach".

The excercise also points to an example in our professor's textbook, that states that what I need is a transformation of the data for Xt and Xt-2, as well as Yt and Yt-2, in order to eliminate the effects of Xt-1/Yt-1. Through this, the original data is replaced with residuals from regressing Xt on Xt-1 and regressing Xt-2 on Xt. The stimates from these regressions are supposed approximate the automatically generated estimated of the PAC on the first lag.

Does anyone have any idea how to do this? What syntax and transformations should I use?