In traditional spectral estimation, the data are often “windowed” by a bell-shaped function to reduce spectral leakage. In the multitaper method, several different window functions are applied, and the results combined. The resulting spectrum has low leakage, low variance, and retains information from the beginning and end of the time series. In addition, statistical testing can take advantage of the multiple spectral estimates. One possible disadvantage is reduced spectral resolution.
Requires evenly spaced data, given in one column. The data are not detrended.
The implementation is based on Mann & Lees (1996) and Fortran code by Michael Mann. It includes a red-noise model based on a “reshaped” spectrum, i.e. after removing and interpolating peaks.
The number of tapers can be set to 3 or 5 for different trade-offs between variance and resolution.
Spectrum type: The two algorithms give very similar results, but the “adaptive” option is recommended by Mann & Lees (1996).
Sample interval: This value only affects the scaling of the frequency axis.
Reshape threshold: This value affects how strong a spectral peak must be to count as a harmonic component, to be removed by the reshaping procedure.
Note: Past reproduces examples in Mann & Lees (1996), but some other implementations seem to give higher levels for the confidence (significance) lines. The reason for this is unknown.
Reference
Mann, M.E. & Lees. J. 1996. Robust estimation of background noise and signal detection in climatic time series. Climatic Change 33:409-445.