Gauss-Vanicek Spectral Analysis of the Sepkoski Compendium: No New Life Cycles

TL;DR

This paper demonstrates that the previously reported 62-million-year life cycle in fossil data is not supported when using the Gauss-Vanicek spectral analysis, highlighting the importance of proper data processing to avoid artificial signals.

Contribution

The paper introduces the Gauss-Vanicek spectral analysis as a robust method for period detection in gapped and noisy data, and shows its effectiveness in verifying or refuting previously claimed cycles.

Findings

No 62 Myr cycle found with Gauss-Vanicek analysis

Data manipulation can artificially inflate spectral signals

Gauss-Vanicek method improves reliability of spectral analysis

Abstract

New periods can emerge from data as a byproduct of incorrect processing or even the method applied. In one such recent instance, a new life cycle with a 62+-3 Myr period was reportedly found (about trend) in genus variations from the Sepkoski compendium, the world most complete fossil record. The approach that led to reporting this period was based on Fourier method of spectral analysis. I show here that no such period is found when the original data set is considered rigorously and processed in the Gauss-Vanicek spectral analysis. I also demonstrate that data altering can boost spectral power up to a nearly 100 percent increase in the signal range, thus introducing artificial, "99 percent significant" periods as seen in the corresponding variance-spectra of noise. Besides geology and paleontology, virtually all science and engineering disciplines could benefit from the approach described here. The main general advantages of the Gauss-Vanicek spectral analysis lay in period detection from gapped records and in straightforward testing of statistical null hypothesis. The main advantage of the method for physical sciences is its use as a field descriptor for accurate simultaneous detection of eigenfrequencies and relative dynamics. Besides analyzing incomplete records, researchers might also want to remove less-trustworthy data from any time series before analyzing it with the Gauss-Vanicek method. This could increase both the accuracy and reliability of spectral analyses in general.