A mixture-type limit theorem for nonlinear functions of Gaussian sequences (Q805059)

The authors consider a Gaussian stationary process \(X_ n\), \(n\in Z\), E \(X_ n=0\), E \(X^ 2=1\), with correlation function \(r(n)=E X_ 0X_ n\) satisfying \(r(n)\approx | n|^{-\alpha}\), \(0<\alpha <1\) (that is \(r(n)| n|^{\alpha}\to 1\) as \(| n| \to \infty)\). Then, the set of all the square integrable random variables measurable with respect to the \(\sigma\)-field generated by the process \(\{X_ n\}\) is denoted by \({\mathcal H}\). It is shown that for a certain class of functions in \({\mathcal H}\) it is possible to obtain a mixture-type limit theorem in which the limiting distribution corresponds to the sum of two random variables, one Gaussian and the other non-Gaussian, represented by a multiple Wiener-Itô integral with the two random variables independent.