Skip to contents

Convert VHAR process to infinite vector MA process

Usage

VHARtoVMA(object, lag_max)

Arguments

object

A vharlse object

lag_max

Maximum lag for VMA

Value

VMA coefficient of k(lag-max + 1) x k dimension

Details

Let VAR(p) be stable and let VAR(p) be \(Y_0 = X_0 B + Z\)

VHAR is VAR(22) with $$Y_0 = X_1 B + Z = ((X_0 \tilde{T}^T)) \Phi + Z$$

Observe that $$B = \tilde{T}^T \Phi$$

References

Lütkepohl, H. (2007). New Introduction to Multiple Time Series Analysis. Springer Publishing.