Main Content

Vector Autoregression Models

Stationary multivariate linear models including exogenous predictor variables

A vector autoregression (VAR) model is a system of simultaneous linear equations that describes the evolution of multiple stationary response series. Equations in the system are functions of constants, time trends, lagged responses, and exogenous predictor variables. For an example of an analysis using VAR modeling tools, see VAR Model Case Study.

To convert your VAR model analysis code from using vgx functions to using the varm object and its object functions, see Convert from vgx Functions to Model Objects.


Econometric ModelerAnalyze and model econometric time series


expand all

varmCreate vector autoregression (VAR) model
estimateFit vector autoregression (VAR) model to data
inferInfer vector autoregression model (VAR) innovations
summarizeDisplay estimation results of vector autoregression (VAR) model
gctestGranger causality and block exogeneity tests for vector autoregression (VAR) models
irfGenerate vector autoregression (VAR) model impulse responses
fevdGenerate vector autoregression (VAR) model forecast error variance decomposition (FEVD)
gctestBlock-wise Granger causality and block exogeneity tests
armairfGenerate or plot ARMA model impulse responses
armafevdGenerate or plot ARMA model forecast error variance decomposition (FEVD)
arma2arConvert ARMA model to AR model
arma2maConvert ARMA model to MA model
vec2varConvert VEC model to VAR model
var2vecConvert VAR model to VEC model
vecmConvert vector autoregression (VAR) model to vector error-correction (VEC) model
simulateMonte Carlo simulation of vector autoregression (VAR) model
filterFilter disturbances through vector autoregression (VAR) model
forecastForecast vector autoregression (VAR) model responses



Create Model

Fit Model to Data

Impulse Response Functions and Granger Causality

Convert Between Models

Generate Simulations or Impulse Responses

Generate Minimum Mean Square Error Forecasts