Problem with AR/Estimation

[zelenaboja]

hi,
I am having a problem obtaining the same results as those int he Damodar Gujarati book. Do not know where is the problem.
I am using the excel in an attachment; E-views 9.5 student version.

in the equation I put "pce c pdi ar(1)"


1. In the book there is -1592,481 for the constant c, and in eviews my result is -661.5021
2. in the book there are only c, pdi and ar(1) variables, but I also get "SIGMASQ", and don't know what it means
3. even though I state to make an LSE, for method I get ARMA Maximum Likelihood (OPG...) and in the book there is only Least Squares method stated

This is what I get:



I want to obtain results exactly as in the book:


P.S. I use only 44 observations, as instructed, not 49. Thank you.

[startz]

EViews has added a new default method for AR(1). The Help system will tell you how to specify the classic method if that's what you want.

[zelenaboja]

I want to obtain values exactly like in the book.
How can I do it? I tried in the options sections....and got this, very close to the book's values...
but can u specify exactly how shoud I set it up...or tell me where to look.
Thanks.

[xprimexinverse]

Hi,

Try the following commands.

Code: Select all

smpl @first 2004
ls(arma=cls, optmethod=legacy) pce c pdi ar(1)


The regression output shown below displays my results and they match those which appear in the book.

ar1reg.png (63.04 KiB) Viewed 4613 times

From my vague recollection of Box and Jenkins' textbook, my understanding is that "cls - conditional least squares" and "Marquardt" technique is the classical way of estimating ARMA models. This seems to be the main option you've omitted.

Hope that helps.

Cheers,
Graeme