VECM forecasts, documentation and bug fix.

[SebKrantz]

Thanks for the package! I would like to generate forecasts from sparse VECM's. Could you add a function to do this? In the meanwhile could you let me know how to do it? It would be great if you could update the documentation insofar as making it clear what the underlying models and the matrices in the output are, e.g. what are the Pi and G matrices? Also How do you determine the order of cointegration? (Johannsen?). There also seem to be some bugs in the code. For example I did not manage to run computeForecasts(spVAR, numSteps = 1), where spVAR is a VAR of lag-order p = 3. I believe the error is found in the following line of computeForecasts:

    tmpData <- matrix(data = t(data[(nr - p + 1):nr, ]), 
      nrow = nc, ncol = numSteps)

Setting numSteps = 1 here removes two columns from the data. That line should read:

    tmpData <- matrix(data = t(data[(nr - p + 1):nr, ]), 
      nrow = nc, ncol = p)

I presume you are no longer actively developing this package, but it still seems the only package to estimate sparse VECM's - so it would be great to have a functional toolset for that. Possibly you could also use the function vec2var from the vars package to convert the sparse VECM to a level VAR representation and enable the use of he methods of the vars package on it.

[svazzole]

Hi,
sorry for the delay answering your question and thanks for the interest in the package.
It is true that I'm not actively developing sparsevar, but it is open to any help or contribution.
I will try to look at the problems you have signaled in the weekend.

Thanks a lot.
Best,
Simone