Noncentral limit theorems and Appell polynomials (Q578730): Difference between revisions

Let \(n\geq 1\) be an integer and \(\{\xi_ i\}^{\infty}_{i=-\infty}\) an i.i.d. sequence of random variables with mean 0 and E \(\xi\) \({}_ i^{2n}<\infty\). Consider the moving average \(X_ k=\sum^{k}_{i=- \infty}c_{k-i}\xi_ i\) with \(c_ k\sim k^{-\beta}L(k)\) as \(k\to \infty\), where \(<\beta <(1+n^{-1})/2\), and L(x) is a slowly varying function as \(x\to \infty\). (Such a moving average is said to exhibit a long-range dependence.) Let \(P_ n\) be the nth Appell polynomial associated with the distribution of \(X_ 0\), which can be defined recursively by \(P_ n'(x)=nP_{n-1}(x)\), \(P_ 0(x)=1\), and E \(P_ n(X_ 0)=\delta_{0,n}.\) It is shown in this paper that \(\sum^{[Nt]}_{k=1}P_ n(X_ k)/N^ HL^ n(N)\) converges weakly as \(N\to \infty\) to the Hermite process \(Z_ n(t)\) in the cadlag space D[0,1], where \(H=1-n(\beta -)\) and \[ Z_ n(t) = n!\int...\int \{\int^{t}_{0}\prod^{n}_{j=1}[(v-u_ j)_+]^{-\beta}dv\}dB(u_ 1)...dB(u_ n) \] where \(-\infty <u_ 1<u_ 2<...<u_ n<t\) and B(\(\cdot)\) is standard Brownian motion.