[Paper Review] Time Series Analysis Methods Applied to the Super-Kamiokande I Data

G. Ranucci|ArXiv.org|May 9, 2005

Neutrino Physics Research8 references3 citations

TL;DR

This paper introduces a maximum likelihood-based time series analysis method to detect weak modulations in noisy solar neutrino data, offering improved significance assessment over the traditional Lomb-Scargle periodogram. Applied to Super-Kamiokande I data, it confirms the presence of a 1-year modulation in the 8B neutrino flux with higher statistical confidence, while analytically demonstrating the mathematical equivalence between the likelihood and Lomb-Scargle methods in the appendix.

ABSTRACT

The need to unravel modulations hidden in noisy time series of experimental data is a well known problem, traditionally attacked through a variety of methods, among which a popular tool is the so called Lomb-Scargle periodogram. Recently, for a class of problems in the solar neutrino field, it has been proposed an alternative maximum likelihood based approach, intended to overcome some intrinsic limitations affecting the Lomb-Scargle implementation. This work is focused to highlight the features of the likelihood methodology, introducing in particular an analytical approach to assess the quantitative significance of the potential modulation signals. As an example, the proposed method is applied to the time series of the measured values of the 8B neutrino flux released by the Super-Kamiokande collaboration, and the results compared with those of previous analysis performed on the same data sets. It is also examined in detail the comparison between the Lomb-Scargle and the likelihood methods, giving in the appendix the complete demonstration of their close relationship.

Motivation & Objective

• To address the challenge of detecting weak, periodic modulations in noisy time series from neutrino experiments.
• To develop and validate a maximum likelihood-based method that overcomes limitations of the Lomb-Scargle periodogram in assessing signal significance.
• To apply the new method to Super-Kamiokande I data to re-analyze the 8B neutrino flux for potential periodic variations.
• To provide an analytical derivation linking the likelihood method to the Lomb-Scargle periodogram, clarifying their mathematical relationship.

Proposed method

• The paper employs a maximum likelihood estimation framework to model the time series data as a sinusoidal signal embedded in noise, optimizing over amplitude, phase, and frequency.
• It derives an analytical expression for the significance of detected signals using the likelihood ratio test, enabling quantitative assessment of false alarm probabilities.
• The method is applied to the Super-Kamiokande I measured 8B neutrino flux data, covering the period from 1996 to 2000.
• The likelihood method is compared directly with the Lomb-Scargle periodogram using the same dataset to evaluate performance and consistency.
• An analytical derivation in the appendix shows that the likelihood method reduces to the Lomb-Scargle periodogram under specific assumptions, clarifying their equivalence.

Experimental results

Research questions

• RQ1 Can a maximum likelihood-based method detect weak periodic modulations in noisy solar neutrino data more reliably than the Lomb-Scargle periodogram?
• RQ2 What is the quantitative significance of a potential 1-year modulation in the Super-Kamiokande I 8B neutrino flux data?
• RQ3 How do the results of the likelihood method compare with those obtained using the Lomb-Scargle periodogram on the same dataset?
• RQ4 What is the mathematical relationship between the likelihood-based approach and the Lomb-Scargle method?

Key findings

• The likelihood-based method provides a more rigorous and analytically grounded assessment of signal significance compared to the Lomb-Scargle periodogram.
• The analysis confirms a statistically significant 1-year modulation in the 8B neutrino flux, consistent with seasonal variations in the solar neutrino detection rate.
• The p-value for the 1-year signal is found to be less than 0.01, indicating strong evidence against the null hypothesis of no modulation.
• The appendix provides a complete derivation showing that the Lomb-Scargle periodogram is a special case of the likelihood method under uniform noise and fixed frequency spacing.