Stat Colloquium [In-Person]: Dr. Antonio Possolo
NIST Fellow and Chief Statistician
Title: Measurement science meets the reproducibility challenge
Abstract: Consensus building exercises conducted in measurement science serve primarily to demonstrate reproducibility of measurement results produced by laboratories measuring the same property independently of one another, thus ensuring the exchangeability of measurements made worldwide in support of commerce, industry, agriculture, science and technology, and medicine.
Inter-laboratory studies in measurement science can deliver a consensus value that is both a more precise and more reliable estimate of the true value of the property of interest than any of the estimates produced by the individual studies, and also serve to demonstrate the participants' measurement capabilities. By pooling evidence gathered in multiple studies, meta-analysis in medicine enhances confidence in conclusions that, if taken one separately from the other, may not be decisive about the relative merits of alternative therapies.
This presentation illustrates and discusses several of the key reproducibility challenges facing inter-laboratory comparisons in metrology and meta-analytic studies in medicine, as they track the truth about the issues that they address.
The illustrative examples include: (1) an inter-laboratory study of the stress required to achieve a specified elongation of natural rubber; (2) a meta-analysis of the effect of rosiglitazone as a potentially aggravating factor for the risk of myocardial infarction; (3) the estimation of the reproduction number for COVID-19 by multiple research groups, and alternative models for pooling their results into a consensus value; and (4) the mutually inconsistent measurement results for the Newtonian constant of gravitation, G, and the steps being taken to achieve reproducibility.
Publications:
(i) A. Possolo (2023) Measurement science meets the reproducibility challenge. Metrologia 60: 044002. DOI 10.1088/1681-7575/acdef7.
(ii) A. Possolo (2023) Tracking Truth through Measurement and the Spyglass of Statistics. Statistical Science 38(4): 655-671. DOI 10.1214/23-STS899.