The RATS Software Forum

Dear Tom　Doan,

I need quick help from you. I have 4 variables. They are non-stationary at level but 1st difference, they are stationary. At level they are not cointegrated. It means, they have not long run relationship. Can you please help me what methods I will employ to estimate these variables.

I am looking your help

With Regard

Ali

Dear Tom　Doan,

Thank you for your reply. I am writing again because I differ from my supervisor suggestion. According to your advice, I take 1st difference data (all variables are stationary at 1st difference) and use OLS for estimation. I also check Diagnostics Statistics such as Spurious Regression, Wald Test and Heteroskedasticity Test. All provide good result.The group of data are used in this analysis are not contigrated. So, my suggestion is that I do not need to do VECM because the varibles are not co-integrated at level but my supervisor suggested me to do VECM.

Now I am looking your suggestion.

I appreciate you for your help

With Kind
Ali

If you have a non-trivial VECM (non-zero coefficient on a linear combination of lagged levels), then that describes a system which has to show cointegration. Based upon your description a VECM is unnecessary.

24 Sept' 2012

Dear Tom　Doan,

Thank you for your great help. I appreciate you and your work. Again I need your quick help regarding the interpretation of VAR model. I have 3 variables (one dependent variable and two independent variables) in VAR Model and 4 lag lengths are appropriate. Four Coefficients (for each independent variable) jointly affect the dependent variable. So, I like to know which coefficient I will explain in my paper because each variable has 4 coefficients or there is any other system to explain. Please let me know.

I am looking your urgent help to solve the problem.

Thanking you

Kind Regards
Ali

Could you please not post the same question twice?

Why are you saying you have a VAR with one dependent variable? That's not a VAR. It sounds more like you have an ARDL model. Whether it's a VAR or an ARDL, the individual coefficients aren't particularly interesting---it's the dynamic responses that matter. It's considered bad form and a waste of space to list individual coefficients in a model like that.

28 Sept' 2012

Dear TomDoan

Thank you for your help and advice. I would be greatful to you if you kindly tell me how do i estimate the above situation. Can you please explain little bit more about ARDL such as how do I estimate ARDL model. Suppose I have 3 variable (Y, X and Z). Y is stationary at level and X and Z is stationary at first difference. At most one is co-integrated. How do I write equation for ARDL to estimate. I am waiting for your help.

Thanking you

With Kinds
Ali

Dear Tom Doam

i appreciate your help all the times. Again I am facing new problem and I need your quick help. We know that when variables at level are non-stationary but 1st difference are stationary and they are found co-integrated, we employ VECM to measure short-run relationship. Now I like to know how do I estimate long-run relationship. For example, as we know that our variables are co-integrated, so we can run multiple regression analysis to measure long-run relationship by using at level data or Johansen system cointegration test, there is nomalized cointegrating coefficients, we can explain this coefficients as a long-run relationship.

Please make me clear and I am waiting for your reply

Thanking you

Ali

That's not a "quick help". There's a huge literature on methods of determining the cointegrating rank and estimating the cointegrating vectors.