EViews.com

Eviews instruction page says - DOLS estimated output doesn't display the coefficients related to extra lead and lag terms that are added. It also said, if one desires to see those coefficients can do so by including these lag/lead terms (short-run dynamics) and then use OLS. I did this - that produces same coeffcieints in the cointegrating equations with same r^2, adj-r^2, se of regression. It produced same residuals. But t-stats were different.
Question: Can you tell me why the differences in t-stats? Is there a way to compute the same t-stats as in DOLS?

Muin

From the manual for DOLS:

The default computation method re-scales the ordinary least squares coefficient covariance using an estimator of the long-run variance of DOLS residuals (multiplying by the ratio of the long-run variance to the ordinary squared standard error).

http://www.eviews.com/help/helpintro.ht ... on.html%23


The problem with trying to match these results is that the OLS estimator doesn't offer an equivalent method as an option. The closest you can do is to save the residuals from the DOLS, and use the long-run variance view on the series to compute the appropriate scaling factor and then scale the OLS standard errors accordingly.

Alternatively, you can change the options on DOLS to compute a full HAC estimator for DOLS, and then compute the OLS equation using the equivalent HAC.

Hi Glenn,

As far I can tell both esimations (DOLS) and corresponding OLS use the same HAC options.

Muin

Are you using the default DOLS coefficient variance estimator? If so, the options may be the same, but it is computing a different estimator of the coefficient variance.