Welch's method has been my go-to algorithm for computing power spectral density (PSD) of evenly-sampled timeseries. I noticed that there are many other methods for computing PSD. For example, in Matlab I see:
- PSD using BurbBurg method
- PSD using covariance method
- PSD using periodogram
- PSD using modified covariance method
- PSD using multitaper method (MTM)
- PSD using Welch's method
- PSD using Yule-Walker AR method
- Spectrogram using short-time Fourier transform
- Spectral estimation
What are the advantages of these various methods? As a practical question, when would I want to use something other than Welch's method?
