Abstract

This study investigates the reliability of the Scopus author identifiers (AUIDs) for research evaluation purposes at different scales. Differences in the scientific performance of researchers across funding bins were first computed using manually cleaned publication portfolios for populations of various sizes (from 20 up to 1,500 researchers per bin, for 5 bins; i.e., for populations from 100 to 7,500 researchers). The differences in performance across bins based on the AUIDs were then compared to those based on the cleaned portfolios (the “ground truth”). To the degree that the findings based on the AUIDs approximate those based on the manual cleaning, the AUIDs can reliably be used in place of manual cleaning. The results show that when comparing groups of fewer than 500 researchers, manual cleaning should be used. With groups of roughly 500 researchers, manual cleaning should still proceed if budget is sufficient. Otherwise, AUIDs can be used providing that care is taken not to overinterpret the findings. With groups of around 1,000 or more researchers, the use of Scopus AUIDs is warranted to decrease study costs. At that scale, the AUIDs provide highly reliable conclusions in terms of direction and size of effect.

See the conference paper here [PDF].

Image credit: Photo by Joanjo Pavon on Unsplash