No label defined (Q4940244)

Let \(w(\vec{x})\), \(\vec{x}\in[\vec 0,\vec T]\), \(\vec T=(T,\ldots,T)\in R_+^{m}\), be a Wiener field, that is \(w(\vec x)\) is a real-valued random field almost sure continuous and vanishing on the coordinate hyperplanes and \(Ew(\vec x)=0\), \(Ew(\vec x)w(\vec y)= \prod\limits_{i=1}^{m}\min(x^{i},y^{i}).\) The authors consider random fields \(\xi(\vec y)\) which have the representation \[ \xi(\vec y)=\int\limits_{[\vec 0,\vec y]^{m}} a(\vec v) d{\vec v} + \int\limits_{[\vec 0,\vec y]^{m}} b(\vec v) dw({\vec v}), \] where \(a(\vec y)\) and \(b(\vec y)\) are random fields such that \[ P\biggl\{ \int\limits_{[\vec 0,\vec T]^{m}} |a(\vec u)|d{\vec u}<\infty\biggr\}=1 \quad\text{and}\quad P\biggl\{ \int\limits_{[\vec 0,\vec T]^{m}} b^2(\vec u)|d{\vec u}<\infty\biggr\}=1. \] The authors prove the Itô formula for \(\xi(\vec y).\) Let \[ \xi(\vec y)=\int\limits_{[\vec 0,\vec y]^{m}} a(\vec u,\xi(\vec u)) d{\vec u} +\;w(\vec y), \] and \(\mu_{\xi}\) and \(\mu_{w}\) be measures generated by \(\xi\) and \(w\). They find assumptions under which the measure \(\mu_{\xi}\) is absolutely continuous with respect to \(\mu_{w}.\) They consider an estimate of the parameter \(\theta\) for \[ \xi_{\theta}(\vec y)=\theta\int\limits_{[\vec 0,\vec y]^{m}} a(\vec u,{\vec\xi}(\vec u)) d{\vec u} +\;w(\vec u). \] The maximum likelihood estimator \(\theta_{T}\) of \(\theta\) has the form \[ \theta_{T}=\int\limits_{[\vec 0,\vec T]^{m}} a(\vec x,{\vec\xi}(\vec x)) d{\vec \xi}(\vec x)\;\biggl(\int\limits_{[\vec 0,\vec T]^{m}} a^2(\vec x,{\vec\xi}(\vec x)) d{\vec x}\biggr)^{-1}. \] The authors find exponential estimates for the probability \(P\bigl\{ T^{m/2}|\theta_{T}-\theta|>R \bigr\}.\) Hypotheses tests for measures generated by \(\xi(\vec y)\) are also considered.