Dear Researchers
When the dependent variable is i(0). can we apply the ARDL model in this case. if no what i do because i am going to estimate the export supply model at dis aggregated level but in the case of some categories the dependent variable is i(0) whereas some regressors are i(1) or some are i(0).
thanks and best regards
Irshad
Hi.
Yes, you can. For all I(0) VAR, for all I(1) cointegration first and for mixed of I(0) and I(1) ARDL.
Thanks.
ARDL is applied for model specifications with lagged independent and dependent variables, i.e. variables are connected over time. Therefore, I see no sense in applying ARDL in case of i(0) for the dependent variable. But it could well be, that an integration test does not show correctly. So try it, and you wll see if you get a reaonable significant lag structure (I expect "no"). I would be glad for i(0), because it would mean a simpler specification (no distributed lags) and more degrees of freedom (making statistics "sharper").
Hi
I agree with Mr. Rainer. there is any problem when "some of" independent variables are I(0) (it`s clear that, any of them should not be I(2)). but using OLS for estimation, when you have some I(1) variables, (and avoiding spurious regression) has this implicit assumption that independent variables that are I(1) has a cointegration relationship with dependent variable that, is I(1). So your dependent variable should be I(1).
best regards
@syed Irshad Hussain question is being misunderstood.He is asking about I(0) dependent variable , not asking about a mix on the right hand side variables.
Muhammad, you are right, but if a variable is dependent from an exogenous variable with different lags, it is, I suppose, unlikely that it is i(0). In any case, for an i(0) dependent variable, it is, at least, unnecessary to apply ARDL, because there is no autoregressivity.
No autoregressivity? Why. Autoregressive refers to lags while stationarity refers to the level of integration, differences here..Isn't it so?
Muhammad, you are right, but I think it rather unlikely that an autoregressive process also including lagged explaining variables will result in a stationary variable. Apart from that, ARDL as a modern method is often used to "save" a weak theoretical model.
If the dependent variable is an I(0), and in a single regression model, you cannot use cointegration test or long run analysis. In this case, you use the short run analysis where all variables are stationary. The ARDL is useful if the dependent variable is not an I(0) and the independent variables have different orders of integration.
The dependent variable has to be I(1), while the independent variables can be a mix of I(0) and I(1) variables. You cannot regress stationary variable on nonstationary variables.
Also see attached paper for more info.
hope this helps
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication