You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
40 2. Gaussian fields<br />
Their diversity shows the power of the abstract approach, in that all can<br />
be treated via the general theory without further probabilistic arguments.<br />
The reader who is not interested in the general Gaussian theory, and cares<br />
mainly about the geometry of fields on R N , need only read Sections 2.2.1<br />
and 2.2.2 on continuity and differentiability in this scenario.<br />
The next two important results are the Borell-TIS inequality and Slepian’s<br />
inequality (and its newer relatives) in Sections 2.3 and 2.4 respectively. The<br />
Borell-TIS inequality gives a universal bound for the tail probability<br />
P{sup f(t) ≥ u},<br />
t∈T<br />
u > 0, for any centered, continuous Gaussian field. As such, it is a truly basic<br />
tool of Gaussian processes, somewhat akin to Chebychev’s inequality in<br />
Statistics or maximal inequalities in Martingale Theory. Slepian’s inequality<br />
and its relatives are just as important and basic, and allow one to use<br />
relationships between covariance functions of Gaussian fields to compare<br />
the tail probabilities and expectations of their suprema.<br />
The final major result of this Chapter is encapsulated in Theorem 2.5.1,<br />
which gives an expansion for a Gaussian field in terms of deterministic<br />
eigenfunctions with independent N(0, 1) coefficients. A special case of this<br />
expansion is the Karhunen-Loève expansion of Section 2.5.1, with which<br />
many readers will already be familiar. Together with the spectral representations<br />
of Section 1.4, they make up what are probably the most important<br />
tools in the Gaussian modeller’s box of tricks. However, these expansions<br />
are also an extremely important theoretical tool, whose development has<br />
far reaching consequences.<br />
2.1 Boundedness and continuity<br />
The aim of this Section is to develop a useful sufficient condition for a<br />
centered Gaussian field on a parameter space T to be almost surely bounded<br />
and/or continuous; i.e. to determine conditions for which<br />
P{sup |f(t)| < ∞} = 1 or P{lim |f(t) − f(s)| = 0, ∀t ∈ T } = 1.<br />
t∈T<br />
s→t<br />
Of course, in order to talk about continuity – i.e. for the notation s → t<br />
above to have some meaning – it is necessary that T have some topology, so<br />
we assume that (T, τ) is a metric space, and that continuity is in terms of<br />
the τ-topology. Our first step is to show that τ is irrelevant to the question<br />
of continuity 2 . This is rather useful, since we shall also soon show that<br />
2 However, τ will come back into the picture when we talk about moduli of continuity<br />
later in this Section.