In time series analysis, if some of the variables are stationary at level, some are at first- difference 1 and second-difference, then which model will be appropriate to check the long run and short run relationship.
Do you mean the independent variables become stationary at different levels? If so, then you can use them at respective levels at which they become stationary.
yas you can using ARDEL model and also you can using VAR model ... Regards
Autoregressive Distributive Lag (ARDL) model is definitely what you need to to check the long run and short run relationship
You might check the paper
Tsay, W.J., The Long Memory Autoregressive Distributed Lag Model and Its Application on Congressional Approval. Electoral Studies, 29, 128-143, 2010
where it deals with the ARDL with stationary or nonstationary long memory regressors and errrors.
Why not run a simple OLS-regression in first differences for all variables, because the differences of stationary variables should be stationary, too?
ARDL is not an economic model, it is just a statistical method (like OLS etc.). Most econometricians seem to concentrate on methods and forget economics. In fact, a well-specified economic model/equation is by far more important. One has to choose the method according to this specification, and I think, one should apply the simplest possible method for the beginning. One should also examine the relations between the independent variables and, maybe, change the specification (transformations of variables) to avoid spurious results. I do not understand, why ARDL is the right method for any economic problem. For example, if according to the economic model/specification past values of the dependent variable have no influence on its present one, it makes no sense to use AR (I would even say: it is a fault!)
ANYON i am agree with you. One of the draw back of the regression or correlation is that if we fitted such model at any kind of data it will given us estimation result on the basis of data untle in relaity there will be no any relation
As I do not know your specification, I it is impossible to give reasonabel advice. Maybe your approach is simply wrong, but this does not mean that it is not valuable, because you should show the results, that everyone knows, that the hypothesis is false. One should not try to "verify" a hopeless hypothesis.
I hope that the answers and recommendations (mostly method-related without knowing what should be analysed) do not mean that the majority of econometricians are not economists.
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series