A central limit theorem for quadratic forms in strongly dependent linear variables and its application to asymptotical normality of Whittle's estimate (Q2638684): Difference between revisions

Let \(X(t)=\sum_{s\in {\mathbb{Z}}}a(t-s)\xi_ s\), \(t\in {\mathbb{Z}}\), where a(t), \(t\in {\mathbb{Z}}\), are real weithts, \(\sum_{t\in {\mathbb{Z}}}a^ 2(t)<\infty\) and \(\xi_ s\), \(s\in {\mathbb{Z}}\), is an i.i.d. sequence with mean 0, variance 1 and finite fourth moment. Let \[ r(t)=\int^{\pi}_{-\pi}e^{itx}f(x)dx \] be the covariance of this linear sequence. Put \[ Q_ N=\sum^{N}_{t,s=1}b(t-s)X(t)X(s)\text{ where } b(t)=\int^{\pi}_{-\pi}e^{itx}\hat b(x)dx,\quad t\in {\mathbb{Z}}; \] \(\hat b(x)\) is a real, even, integrable function on \([- \pi,\pi]\). Denote \[ B_ N=(b(t-s))_{t,s=\overline{1,N}},\quad R_ N=(r(t-s))_{t,s=\overline{1,N}}. \] If \[ Tr(R_ NB_ N)^ 2/N\to (2\pi)^ 3\int^{\pi}_{-\pi}(f(x)\hat b(x))^ 2dx<\infty \] then \[ (Q_ N-E Q_ N)/\sqrt{N}\to^{{\mathcal D}}N(0,\sigma^ 2), \] where \(\sigma^ 2\) is specified. This generalization of results of \textit{F. Avram} [ibid. 79, No.1, 37-45 (1988; Zbl 0648.60043), and \textit{R. Fox} and \textit{M. S. Taqqu} [ibid. 74, 213-240 (1987; Zbl 0586.60019)] for Gaussian variables is applied to prove the asymptotic normality of Whittle's estimate.