Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser
Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser
Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
2.2. LINEAR PROCESS 8<br />
Finally, I provide examples using geophysical series.<br />
2.2 Linear Process<br />
It should be stressed that any wide sense stationary random process can be rep-<br />
resented <strong>via</strong> the superposition <strong>of</strong> a deterministic part sn plus a non deterministic<br />
part xn (random or stochastic part),<br />
zn = sn + xn. (2.1)<br />
Wold decomposition theorem (Wold, 1965; Ulrych and Sacchi, 2005), in addition<br />
states that the non deterministic part can be represented as a filtered sequence <strong>of</strong><br />
white noise<br />
∞<br />
xn = giεn−i<br />
i=1<br />
(2.2)<br />
where g0 = 1, ∞ i=1 |gi| 2 < ∞ (finite length impulse response) and εn represents the<br />
white noise which is uncorrelated with sn. I can change the equation above by a<br />
finite length discrete general linear model<br />
N<br />
xn = giεn−i<br />
i=1<br />
(2.3)<br />
where εn is the innovations process and xn depends on its past input values (there-<br />
fore it is casual).