This question was asked before but has not been answered so far.

I am interested in impulse response analysis in a VAR with cointegration. My sample is not large enough to apply the approach by Toda and Yamamoto [1995].

In Lütkepohl‘s 2007 edition of “Introduction to Multivariate Time Series Analysis“, Phillips‘ 1995 fully modified VAR is mentioned as an alternative to Toda and Yamamoto. Unfortunately, Lütkepohl does not provide any further details.

Does someone know how to estimate a fully modified vector autoregression (FM-VAR) in any common computer package?

As far as I understand, FM-VAR extends fully modified least squares (FMOLS) to VARs.

Of course, any link to other procedures that lead to valid inference (at least in the asymptotic sense) when cointegration and different orders of integration {I(0) up to I(2)} are present, will be much appreciated.

Cheers!