Question

I’m using the vars package and want to predict some values from the calculated models:

# Get the model
x1 <- rnorm(15)
y1 <- x1 + rnorm(15)
trainFrame=data.frame(x1,y1);
model=VAR(trainFrame, p=3);
pr1=predict(model, trainFrame);
# Forecast values with new data
x2 <- rnorm(15)
y2 <- x2 + rnorm(15)
newFrame=data.frame(x2,y2);
pr2=predict(model, newFrame);

Comparing the two prediction vectors pr1 and pr2 shows that they are the same.
How can I get the actual forecast values and not again the forecasts from the training data?

Answer 1

Here you call the predict method for objects with class attribute varest.

predict(object, …, n.ahead = 10, ci = 0.95, dumvar = NULL)

n.ahead forecasts are computed recursively for the estimated VAR.

No need to give the training Frame to predict.

  pr2=pr1=predict(model)