256. VAR

Vector autoregression

i.e.,

Each component depends on lags of all series.

Parameters: ( each), (), ()
Orders: , (series)

Example:

Given

  • Order: ,
  • Coefficient matrix:

  • Intercept:
  • Two series stacked into a vector :
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
12 10 8 11 14 12 9 13 16 14 11 15 18 16 13 17
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
8 9 7 6 10 11 9 8 12 13 11 10 14 15 13 12

Step 1 — formula

Substitute into the VAR recursion:

Forecast (set ):

Componentwise (write out the matrix-vector product):

Innovation:

Step 2 — apply at

Plug in , , :

Step 3 — iterate

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17