Review of Methods of Power-Spectrum Analysis as Applied to Super-Kamiokande Solar Neutrino Data

TL;DR

This study compares six power-spectrum analysis methods on Super-Kamiokande solar neutrino data, showing that more comprehensive methods reveal higher significance peaks, clarifying discrepancies in previous analyses.

Contribution

It systematically evaluates how different information levels in analysis methods affect the detection of significant periodicities in solar neutrino data.

Findings

More informative methods increase peak significance

Results align with most previous analyses

Inability to replicate Koshio's recent findings

Abstract

To help understand why different published analyses of the Super-Kamiokande solar neutrino data arrive at different conclusions, we have applied six different methods to a standardized problem. The key difference between the various methods rests in the amount of information that each processes. A Lomb-Scargle analysis that uses the mid times of the time bins and ignores experimental error estimates uses the least information. A likelihood analysis that uses the start times, end times, and mean live times, and takes account of the experimental error estimates, makes the greatest use of the available information. We carry out power-spectrum analyses of the Super-Kamiokande 5-day solar neutrino data, using each method in turn, for a standard search band (0 to 50 yr-1). For each method, we also carry out a fixed number (10,000) of Monte-Carlo simulations for the purpose of estimating the significance of the leading peak in each power spectrum. We find that, with one exception, the results of these calculations are compatible with those of previously published analyses. (We are unable to replicate Koshio's recent results.) We find that the significance of the peaks at 9.43 yr-1 and at 43.72 yr-1 increases progressively as one incorporates more information into the analysis procedure.