Model uncertainty and reference value of the Planck constant

TL;DR

This paper introduces statistical models to analyze and reconcile different measurements of the Planck constant, accounting for heterogenous uncertainties and improving the reliability of its reference value.

Contribution

It develops a novel approach using model selection and averaging to incorporate model uncertainty into the estimation of the Planck constant.

Findings

Identifies a consistent reference value for the Planck constant.

Demonstrates the importance of accounting for heterogeneous uncertainties.

Provides a framework for integrating model uncertainty into fundamental constant measurements.

Abstract

Statistical parametric models are proposed to explain the values of the Planck constant obtained by comparing electrical and mechanical powers and by counting atoms in Si 28 enriched crystals. They assume that uncertainty contributions -- having heterogeneous, datum-specific, variances -- might not be included in the error budgets of some of the measured values. Model selection and model averaging are used to investigate data consistency, to identify a reference value of the Planck constant, and to include the model uncertainty in the error budget.