Thursday, January 23, 2014

An ARDL Add-in for EViews

My posts on ARDL models and bounds testing (here and here) have certainly been popular. So, I was really pleased to see that Yashar Tarverdi has produced an "Add-In" for the EViews package that makes this type of econometric analysis somewhat easier.

You can download the the add-in program and its installer here. The add-in is called "ARDLbound", and it largely automates the key steps associated with bounds testing using an ARDL model.



The main limitation of the program, as far as I can see, is that it allows for only one explanatory variable, and of course often such models include several "X" variables. None the less, this is a great resource, so a big "thank you" to Yashar for putting this together!

Here's an illustration of what you get with the ARDLbound add-in.

My dependent variable is y and the regressor variable is x.The unit root testing is going to be undertaken with the ADF test; I'm going to allow for up to 6 lags of the variables in the model; and the lag specification is going to be based on minimizing the SIC values.


The program chooses an ARDL(1,1) specification:


Here are the various statistics associated with every possible lag specification:



And here is the preferred ARDL model:

Although this is the preferred specification on the basis of minimizing SIC, in this instance not all of the variables in the ARDL model are statistically significant. But this is just a "pretend example", to show you what to expect when you use this add-in.

Now - off to the sandpit, and play nicely!


© 2014, David E. Giles

14 comments:

  1. I have to thank David and Gareth from Eviews who helped me for debugging this add-in.
    Yashar

    ReplyDelete
    Replies
    1. And your effort saves much of the precious times of researchers interested in time series econometrics generally and specifically in cointegrated and integrated time series.

      Delete
    2. the link don't work , can you help me please te get this add please

      Delete
    3. Yes it DOES work - I have just checked it. It is: http://www.eviews.com/Addins/addins.shtml

      Delete
  2. First thank you for what you are posting. How to regress an Ardl through addins in Eviews when our independent variables are more than two?

    ReplyDelete
    Replies
    1. The person who wrote this add-in didn't allow for more than 2 variables. You'll have to set up the model manually, as in my post.

      Delete
  3. the ardl bounds add don't work, can you help me please

    ReplyDelete
    Replies
    1. Yes it does - you have to have EViews installed of course. Also keep in mind I didn't write it - questions should go to the author, Yashar Tarverdi.

      Delete
  4. Dear Sir I would like to run nonliner ARDL model ...Is there any Eviews Package for Nonliner ARDL modeling?

    ReplyDelete
  5. hello sir, can we take 7 lags for dependent variable when we have only 41 yearly observations?

    ReplyDelete
    Replies
    1. Ashfaq - you can do so, but keep in mind that this will reduce your degrees of freedom, and this in turn may reduce the quality of your inferences.

      Delete
  6. Dear Mr. Giles,
    thanks a lot for your work here on the blog.

    I have a question regarding the second version of the ARDLbounds add-in for eviews 7:

    When selecting "No autocorrelation in Resids" as criterion, is column F, i.e. "P of Wald test" the appriopriate column to look at? And if so, is the Null no autocorrelation or autocorrelation?

    Thanks in advance, Philipp

    (I also asked Yashar Tarverdi, the writer of the add-in, per mail. I will post the answer here if he replies. I only thought that maybe you are faster :) )

    ReplyDelete
    Replies
    1. I suggest you post this in the EViews forum: http://forums.eviews.com/

      Delete