R Statistics Cookbook

How it works...

Essentially, VAR models are the multivariate extensions of AR models. Each one of the equations is estimated via ordinary least squares (OLS). Each one of these equations will contain a total amount of coefficients: number of variables x lags + trend + mean.

The advantage of VAR models is that we can incorporate several variables, thus increasing the predictive power of the model compared to fitting each model individually. But this should be done with some care, because the more irrelevant variables that we add, the wider the confidence intervals will be for the predictions.