Database Reference
In-Depth Information
Figure 8.6
ACF plot of a simulated MA(3) time series
To understand why this phenomenon occurs, it is useful to examine
Equations 8.10
through
8.14
for an MA(3) time series model:
Because the expression of shares specific white noise variables with the
expressions for through , inclusive, those three variables are correlated
to . However, the expression of in
Equation 8.10
does not share white noise
variables with in
Equation 8.14
. So the theoretical correlation between and
is zero. Of course, when dealing with a particular dataset, the theoretical
autocorrelations are unknown, but the observed autocorrelations should be close
to zero for lags greater than q when working with an MA(q) model.