Multitaper.jl: A Julia package for frequency domain analysis of time series
Charlotte Louise Haley, Christopher J. Geoga · The Journal of Open Source Software · 2020
Spectral analysis is widely used to infer physical mechanisms for underlying process dynamics from a realization of a stationary time series.The multitaper method is a nonparametric technique for estimating the power spectrum of a discrete time series that simultaneously controls bias and variance of the estimator by premultiplying the data by a set of orthogonal sequences-discrete prolate spheroidal sequences that are optimally concentrated in both time and frequency.These have the effect of stabilizing the bias and variance of the spectrum estimator (Thomson, 1982).While multitaper codes have been introduced in multiple languages, including Julia, the Multitaper.jlpackage that we present offers functionality beyond univariate and bivariate time series analysis and provides routines for a number of (selected) research-level topics not found in other packages.