257. VMA
Vector moving average
i.e.,
Each component is a linear combination of past shocks across all series.
Parameters: ( each), (), ()
Orders: , (series)
Example:
Given
- Order: ,
-
Coefficient matrix:
- Intercept:
- Initial innovation:
- Two series stacked into :
| 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 VMA recursion:
Forecast (set ):
Componentwise:
Innovation:
Step 2 — apply at
Plug in , , :
Step 3 — iterate
Each row uses the previous innovation . Values rounded to 4 decimal places at every step (computation continues from the rounded value).
| 1 | ||||
| 2 | ||||
| 3 | ||||
| 4 | ||||
| 5 | ||||
| 6 | ||||
| 7 | ||||
| 8 | ||||
| 9 | ||||
| 10 | ||||
| 11 | ||||
| 12 | ||||
| 13 | ||||
| 14 | ||||
| 15 | ||||
| 16 |