Hi guys,
A simple thing in the book called "applied time series modelling and
forecasting" (by harris and sollis) bothered me so I would like to ask
to the group:
Consider an AR(3) process like:
(1-aL-bL^2-cL^3)y(t)=u(t).
And suppose that there are two unit roots so that we write the equation
as:
(1-dL)(1-L)(1-L)y(t)=u(t).
The authors say that, if |d|>1, then the second difference of
y(t)(i.e., (1-L)(1-L)y(t)=Delta^2y(t)) will be stationary. Now, my
question is, why |d|>1? Should not it be |d|<1?
Can this be a mistake? If not, why it is bigger than one?
Thank you in advance
P.S: By Delta, I mean the triangular sign which indicates difference,
i.e., y(t)-y(t-1)=deltay(t)).