I am attempting to build a multivariate DOLS model. More specifically, it is model that is already being used where I work, but I am trying to replicate the model in R. Currently, the model is used via eViews.
This is not a programming question. My problem, however, is that when the model is run in eViews, I cannot fully understand what is going on in the backend within eViews sufficiently to replicate the results via a dynamic model in R (although I'm getting close results).
In eViews, the multivariate model uses the following options:
- Dynamic model (the DOLS option)
- Cointegrating regression
- 3 fixed lags, 0 fixed leads
- Constant trend component
dynlm package in R, I have tried building various differenced models, saving the residuals, re-running the undifferenced regression with the residuals, incorporating lags 1 thru 3, etc. And that's just the beginning.
Using that criteria, can someone with a knowledge of eViews please take me thru the steps (not in R, but statistically) of how eViews is building that type of model?
I really want to understand how this process works, to where I can replicate the results on my own.