A law of the iterated logarithm for an estimate of frequency (Q1819457): Difference between revisions

A single theorem is proved. This gives a form of law of the iterated logarithm for \({\hat \omega}\) an estimate of \(\omega_ 0\) in the model \[ y(t)=\rho_ 0 \cos (\omega_ 0t+\phi_ 0)+x(t). \] The process x(t) is stationary with zero mean and finite variance and is weakly mixing. The estimate \({\hat \omega}\) is the value of \(\omega\) which maximises the periodogram \(q(\omega)=| (1/T)\sum^{T}_{1}y(t)e^{it\omega}|^ 2\) and the result is that \[ \limsup_{T\to \infty}| T^{3/2}(2 \log \log T)^{-1/2}\{48\pi f(\omega_ 0)/\rho^ 2_ 0\}^{-1/2}({\hat \omega}-\omega_ 0)| =1\quad a.s. \] where f(\(\omega)\) is the spectral density of x(t). Some additional conditions on x(t) are also required.