Frequentist model comparison tests of sinusoidal variations in measurements of Newton's gravitational constant

Abstract

Anderson et al have claimed to find evidence for periodic sinusoidal variations (period=5.9 years) in measurements of Newton's Gravitational constant. These claims have been disputed by Pitkin. Using Bayesian model comparison, he argues that a model with an unknown Gaussian noise component is favored over any periodic variations by more than $e^{30}$. We re-examine the claims of Anderson et al using frequentist model comparison tests, both with and without errors in the measurement times. Our findings lend support to Pitkin's claim that a constant term along with an unknown systematic offset provides a better fit to the measurements of Newton's constant, compared to any sinusoidal variations.