Hi Folmer,
When we convert a VAR to VEC, using VEC = var2vec(VAR), we assume that the VAR coefficients are known, so that the VEC coefficients do not carry standard errors. If VAR coefficients are estimated with standard errors, we may assume normality and consider generating random draws centered on the point estimator. Then we set up a FOR loop and apply var2vec(...) for each coefficient draw, the collection of VECs provides a distribution for the transformed coefficients.
If we just want to estimate a VEC, why not fit the model directly? Run [h,pValue,stat,cValue,mles] = jcitest(Y), the fifth output struct “mles” contains the estimated VEC coefficients.
Thank you.
Hang Qian
