I have auto-correlation functions and cross-correlation function estimated from a multivariate dataset (but I don’t have the data – only the ACF/CCF). Does anyone know if there exist functions in R for estimating the parameters of a VAR(p) model from this? Or any advice for how to calculate these from the available information? I need this to be able to simulate from an AR-model.

For example, using the mAR-package, I can simulate using the mAr.sim function if I have the A and C matrices of VAR coefficients and the residual covariance matrix. Is it possible to get these from the ACF and CCF functions?

I have seen that there is a function that can do this in the univariate case, acf2AR in the stats package, but haven’t found anything for multivariate models. Can anyone help – either with reference to packages that can do this or any advice on how to implement this from scratch. Also, is it really possible to estimate the residual covariance matrix without the data?