Characteristic correlation time as estimate of optimum filter length in Maximum Entropy Spectral Analysis
A. Rovelli, Angelo Vulpiani · Geophysical Journal International · 1983
An original conceptually simple criterion for the choice of filter length M for Maximum Entropy Spectral Analysis based on the correlation time-scale of data set is proposed.The following formula is obtained from the autocovariance estimates @tThe application of the proposed criterion to various digital geophysical records has provided satisfactory results.The method is well suited to geophysical signals because they often have a remarkable Markovian character in the broader sense, i.e. the autocovariances have values decreasing roughly exponentially as the time lags increase.Computer simulations on mathematically defined time series confirm the reliability of this criterion.