I am a final year student and I would be glad to get as many helps as I could. I have utlized a lot of Youtube vidoes to learn about the ARDL model estimation but I am now confused. This is because different videos provided different approaches, so I am not longer sure of the step by step to doing it.
For instance, Crunch econometrix said to estimate the ARDL model, then afterwards, do the bounds test to check for cointegration. If cointegration exists, then I would need to estimate an ECM model using least squares by manually entering the equation that would contain the ECM into Eviews. On the other hand,if cointegration doesn't exist, then a short-run ARDL using the differenced variables would be estimated using Least squares( I also would manually enter the equation into eviews, but differenced and with lags). I don't know how correct this procedure is especially since other videos I watched didn't have this.
But I am bothered given the fact that nothing is said of the coefficients obtained for the variables and their lags after estimating the first ARDL (I think it is also called the unrestricted error correction model, right?), especially as regards how to interpret them or what their relevant is to the study. I use EViews 12 which provides the coefficient diagnostics. The "Long run form and bounds test" provides the conditional error correction model. what are the relevance of the estimates provides for the coefficients and their lags and how do you interpret them?
Then there is the "error correction form" feauture that provides the ECM regression,which also reveals some estimations for the difference of the variables (with only lags). What is the relevance of these estimates and how do you interpret them?
I would be cautious when trying to learn from youtube videos. At least you should check anything you get from that source with your textbooks.
The link below is a description of the implementation of ARDL in STATA. The introduction sets out the underlying theory. The basic underlying idea depends on using OLS in dynamic equations in levels of the variables. |This has been used for more than 50 years - long before cointegration. One of the ecm expressions can not be estimated because it is non-linear in the parameters. The use and the interrelationships between the three equations are explained in the paper.
https://www2.econ.tohoku.ac.jp/~PDesign/dp/TUPD-2022-006.pdf
You should be able to implement the routines and examples in the paper in the ARDL routines in any other software.
The application of the bounds test in the ARDL routine requires
How many of the YouTube videos set out these requirements?
If you are doing ARDL bounds tests I would not be concerned about the values or significance of the coefficients on the lagged differences. The main interest is in the estimates of the long-term relationship.
Rotimi Ayoade Ogunjumo
Your EVIEWS suggestion is a good alternative to the STATA note that I suggested. I think that parts 1and 2 of the blog are also very important. I would have some doubts about the use of ARDL in his example 2. Surely there is feedback between the 6-month bond and the 1 and 3 month bonds. If so, should he not be using a VECM. The relevant theory is covered in the earlier parts.
Here are the steps to run an ARDL model from start to finish:
To run an ARDL model from start to finish, you can follow the steps below:
To run an ARDL model from start to finish, you can follow these steps:
In summary, to run an ARDL model from start to finish, you can collect the data you need for your model, open the Eviews program, load the data you want to use for your model, click on the “Quick” menu and select “Estimate Equation”, select the variables you want to include in your model, click on the “ARDL” button to specify that you want to run an ARDL model, specify the order of integration for each variable in your model, and click on the “OK” button to run the model.
I think that it will be very simple for you if you use the grtl software instead with the downloaded data
Chuck A Arize Sir, is there any minimum number of observations required while using ARDL? I have 23 years of data. Can I use ARDL?
You can certainly try if you have a relatively small number of variables and not many lags. Think of the number of coefficients you are estimating and the number of observations you will lose because of the lags. It is possible that the routine will run and produce coefficients that are not significant. This does not mean that they are zero but that you have insufficient data to reject the null.
Using the ARDL system to "estimate" something without an understanding of the points I have already raised is a recipe for spurious results. ARDL is very easy to use with various econometric packages but possibly for that reason is often abused.
- Badges
- Science topic
- Similar topics
- Thesis Research