Large deviations for fields with stationary independent increments (Q1343508): Difference between revisions

Consider a field \((\{x(t),t \subset R_ +^ d\}, D)\) with stationary independent increments with \(X(t) = 0\) for \(t\) on the axes and define \(Z_ n(t) = X(nt)/(n_ 1, \dots, n_ d)\) where \(n = (n_ 1, \dots, n_ d) > 0\) and \(nt = (n_ 1t_ 1, \dots, n_ dt_ d)\). The probability distribution \(P_ n\) of \(\{Z_ n(t), t \in [0,1]^ d\}\) can be considered to be a probability measure on \(D[0,1]^ d\) endowed with Skorokhod topology. Define \(J(a) : = \sup_{\theta \in R^ 1} (a \theta - \log \varphi (\theta))\) where \(\varphi (\theta) : = E e^{\theta X (1,1, \dots, 1)}\) which is assumed to be finite for \(\theta\) in the neighbourhood of 0. The author asserts that if \(\lim_{| a | \to \infty} (J(a)/ | a |) = \infty\), then \(\{P_ n\}\) satisfies the large deviation principle on \(D[0,1]^ d\). Also, if \(\{X(t), t \in R^ d_ +\}\) has no Gaussian component, then \(P_ n\) can be considered to be a probability measure on the space of finite signed Borel measures on \([0,1]^ d\) endowed with the weak convergence topology, on which \(\{P_ n\}\) satisfies the large deviation principle. Some general related results have also been proved.