Power-spectrum analysis of Super-Kamiokande solar neutrino data, taking into account asymmetry in the error estimates

Abstract

The purpose of this article is to carry out a power-spectrum analysis (based on likelihood methods) of the Super-Kamiokande 5-day dataset that takes account of the asymmetry in the error estimates. Whereas the likelihood analysis involves a linear optimization procedure for symmetrical error estimates, it involves a nonlinear optimization procedure for asymmetrical error estimates. We find that for most frequencies there is little difference between the power spectra derived from analyses of symmetrized error estimates and from asymmetrical error estimates. However, this proves not to be the case for the principal peak in the power spectra, which is found at 9.43 yr-1. A likelihood analysis which allows for a "floating offset" and takes account of the start time and end time of each bin and of the flux estimate and the symmetrized error estimate leads to a power of 11.24 for this peak. A Monte Carlo analysis shows that there is a chance of only 1% of finding a peak this big or bigger in the frequency band 1 - 36 yr-1 (the widest band that avoids artificial peaks). On the other hand, an analysis that takes account of the error asymmetry leads to a peak with power 13.24 at that frequency. A Monte Carlo analysis shows that there is a chance of only 0.1% of finding a peak this big or bigger in that frequency band 1 - 36 yr-1. From this perspective, power spectrum analysis that takes account of asymmetry of the error estimates gives evidence for variability that is significant at the 99.9% level. We comment briefly on an apparent discrepancy between power spectrum analyses of the Super-Kamiokande and SNO solar neutrino experiments.