Cameron, M.A. (1983, ASQC and the American Statistical Association) CSIRO, Australia
The problem of comparing the outputs of several time series recorders when each receives the same signal together with stationary, additive noise is considered. When the signal is stationary or is the sum of stationary and periodic components, methods of estimating some parameters using frequency domain methods are given. The estimates are obtained by maximizing an approximate Gaussian likelihood but do not require Gaussian data. The estimation is particularly simple if the spectra of the different error processes are equal. Asymptotic properties of the estimators are presented here with an application of the methods. A small simulation study shows that the asymptotic theory is a reasonable approximation for this application.
Time series,Statistical methods,Fourier analysis,Calibration,Gauges