Greetings Everyone,
Please after updating my EViews which is a student version, I noticed that after carrying out the ARDL model estimation, the static of fixed variables are removed from the long run estimation.
Further, when I tried to include the error correction term (ECT) in the short run equation, it turns out to be not significant. However, before the update which I did yesterday, all this issue were not there. In fact, I did run some of my old results from the ECT which were significant in the older version, but the results were not significant.
Please, can the EViews team assist me with this issue?
Regards,
Vincent
Issue with ARDL model after the latest EVIEWS update
Moderators: EViews Gareth, EViews Moderator
-
EViews Mirza
- Posts: 80
- Joined: Sat Apr 22, 2017 8:23 pm
Re: Issue with ARDL model after the latest EVIEWS update
Dear Vincent,
While ARDL has been a very popular feature in EViews, several users have rightly reported that our earlier implementation, whilst matching prevailing implementations, was not entirely conforming to theory. Our latest patch has corrected for these differences and the latest implementation is correct.
In this regard, some things that did not match the theoretical model were the inclusion of static variables in the long-run equation. Naturally, the long run equation consists only of those variables that converge to equilibrium if cointegration is indeed possible in the system. For this reason, static variables are not part of this equation as they are part of the short-run dynamics. Also, we have reworked the way constrains on the constant and trend term enter the long-run equilibrium equation, and the F-bounds test statistic is now being computed correctly.
Given that these are indeed fundamental changes to the way ARDL estimation is being conducted, it is entirely possible that your ECT is no longer significant. Moreover, please note that the p-values reported in the error-correction model are not the usual p-values from a t-distribution, and must be interpreted in the context of non-standard Brownian motion distributions.
We will be releasing a second and third installment of our ARDL blog series soon that will delve into more detail. Please look out for this when it comes out. In the meantime, I invite you to read our first installment here:
http://blog.eviews.com/2017/04/autoregressive-distributed-lag-ardl.html
Best regards
While ARDL has been a very popular feature in EViews, several users have rightly reported that our earlier implementation, whilst matching prevailing implementations, was not entirely conforming to theory. Our latest patch has corrected for these differences and the latest implementation is correct.
In this regard, some things that did not match the theoretical model were the inclusion of static variables in the long-run equation. Naturally, the long run equation consists only of those variables that converge to equilibrium if cointegration is indeed possible in the system. For this reason, static variables are not part of this equation as they are part of the short-run dynamics. Also, we have reworked the way constrains on the constant and trend term enter the long-run equilibrium equation, and the F-bounds test statistic is now being computed correctly.
Given that these are indeed fundamental changes to the way ARDL estimation is being conducted, it is entirely possible that your ECT is no longer significant. Moreover, please note that the p-values reported in the error-correction model are not the usual p-values from a t-distribution, and must be interpreted in the context of non-standard Brownian motion distributions.
We will be releasing a second and third installment of our ARDL blog series soon that will delve into more detail. Please look out for this when it comes out. In the meantime, I invite you to read our first installment here:
http://blog.eviews.com/2017/04/autoregressive-distributed-lag-ardl.html
Best regards
Who is online
Users browsing this forum: No registered users and 70 guests