Information Criteria

In the package there is the possibility to perform model selection (i.e. the VAR order) using three information criteria: AIC, BIC and Hannan-Quinn.

As an example we generate a VAR(3) model

set.seed(42)
sim <- simulateVAR(N = 30, p = 3, sparsity = 0.15)

and then we estimate four different models on the same dataset.

fit1 <- fitVAR(sim$series, p=1, foldsIDs=TRUE)
fit2 <- fitVAR(sim$series, p=2, foldsIDs=TRUE)
fit3 <- fitVAR(sim$series, p=3, foldsIDs=TRUE)
fit4 <- fitVAR(sim$series, p=4, foldsIDs=TRUE)

We call the function informCrit() passing as argument a list containing all the fitted models.

informCrit(list(fit1,fit2,fit3,fit4))
         AIC        BIC  HannanQuinn
1   7.648623 19.1004142  12.25763398
2  -7.086968 10.6893662   0.06748605
3 -16.673240 -1.3055841 -10.48820792
4 -15.957483  0.7201561  -9.24522116