GriffithsSpectrum
Normalized power spectrum estimation using Griffiths adaptive filters.
Usage
Use to generate a high-resolution periodogram for cycle analysis. Best visualized as a heatmap to identify and track multiple market cycles simultaneously.
Background
The Griffiths Spectrum is an adaptive spectral estimation method that provides higher resolution than a standard DFT for short data segments. It fits an all-pole model to the signal using an LMS algorithm, allowing for instantaneous frequency measurement without the windowing artifacts of FFT-based methods.
Parameters
lower_bound(default: 18): Lower period boundupper_bound(default: 40): Upper period boundlength(default: 40): LMS filter length
Formula
[ Pwr(P) = \frac{0.1}{(1 - \sum coef_i \cos(2\pi i/P))^2 + (\sum coef_i \sin(2\pi i/P))^2} ] [ Pwr_{norm}(P) = \frac{Pwr(P)}{\max(Pwr)} ]