Cochrane Orcutt method - estimate GLS

[naruto]

hi people, i am stuck with eviews trying to estimate using cochrane orcutt method instead of the AR(1) method (i am following the eviews study guide and there is instruction given)

After calculating the new residual series, i am not sure how to rerun the equations to the point at which the estimated rho does not change by a lot.

after which the next step is to convert the constant to get the final equation log(xxx)=xxx + xxx*(log(xx) .......

can someone enlighten me as in how to rerun the equations - pls find attached. i ran the same eq a few times at the command window but i didnt see the pho change any where ......

[startz]

I think you've forgotten to quasi-difference the original equation with new values of the serial correlation coefficient. However, unless you have very few observations or are in an unusual situation you'll find using AR(1) and iterated Cochrane-Orcutt to be essentially identical.

[naruto]

hi startz, thanks. are you able to list out the steps to "quasi-difference the original equation with new values of the serial correlation coefficient".

I tried to do that but didn't work out - thanks

yes, i am aware the results most likely is similar if i use AR(1) but i want to compare these 2 tests - probably look at the standard error. i've heard AR(1) is prefered to Cochraine Orcutt though they give the same results ..so just want to see for myself the stats that make AR(1) the preferred choice