Ramsey reset and AR MA terms in LS regression

[bparksb]

Eviews allows AR(1) MA(1) estimation and Ramsey Reset test with those terms.

I can not reproduce the test statistic or regression by hand.

Question is WHAT is used for the FITTED terms in the Ramsey test with the original equation with AR and/or MA terms.

I know:
1. If you use an AR(1) term, the predicteds that are used in the 'Ramsey' equation have some value for observation 1 - I suspect observation 1 is replaced by 0. Known by the mean of the dependent variable - identical with 0 in the first observation of the predicted.
2. I tried using FORECAST to produce both a dynamic and static predicted, both with and without the first observation replaced by 0. None of the four cases produces the same regression displayed for the Ramsey reset equation.

I am using Eviews 7.1 Nov 3 2010 build

My test data set is http://econ413.wustl.edu/testramsey.wf1

The test equation is aara and the Ramsey result from that regression is table aara_table
I used forecast on aara to produce
crime_ratefdwith dynamic with first obs = 0
crime_ratefswith static with first obs = 0
crime_ratefdwout dynamic with first obs = NA
crime_ratefswout static with first obs = NA

The four 'by hand' Ramsey tests are:
aara_h_dwith
aara_h_swith
aara_h_dwout
arra_h_swout

None of those four agree with the tabled Ramsey from the first equation.

NOTE: I used a simple one variable equation as a test, and I used an undated data set to keep things as simple as possible.

Please help - I have about 5 hours invested in this problem.

Bob

[EViews Gareth]

I didn't follow what you did for your manual calculations, but here's how to calculate it manually:

Code: Select all

aara.makeresid resid01 series fit = crime_rate - resid01 smpl @first @first fit = crime_rate smpl @all equation manual.ls crime_rate c ex0 fit^2 ar(1)

[bparksb]

THANKS that explains it!
Bob