04.01.2013 Views

Springer - Read

Springer - Read

Springer - Read

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Figure 10-2<br />

The sample correlation<br />

functions of the estimated<br />

residuals from the<br />

model fitted in Example<br />

10.1.1. Series 1 is { ˆZt }<br />

and Series 2 is { ˆWt }.<br />

10.1 Transfer Function Models 337<br />

10.1.1 Prediction Based on a Transfer Function Model<br />

When predicting Xn+h,2 on the basis of the transfer function model defined by (10.1.1),<br />

(10.1.4), and (10.1.7), with observations of Xt1 and Xt2, t 1,...,n, our aim is to<br />

find the linear combination of 1,X11,...,Xn1,X12,...,Xn2 that predicts Xn+h,2 with<br />

minimum mean squared error. The exact solution of this problem can be found with<br />

the help of the Kalman recursions (see TSTM, Section 13.1 for details). The program<br />

ITSM uses these recursions to compute the predictors and their mean squared errors.<br />

In order to provide a little more insight, we give here the predictors ˜PnXn+h<br />

and mean squared errors based on infinitely many past observations Xt1 and Xt2,<br />

−∞

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!