Unravelling the performance of individual scholars: use of Canonical Biplot analysis to explore the performance of scientists by academic rank and scientific field

Abstract

Individual research performance needs to be addressed by means of a diverse set of indicators capturing the multidimensional framework of science. In this context, Biplot methods emerge as powerful and reliable visualization tools similar to a scatterplot but capturing the multivariate covariance structures among bibliometric indicators. In this paper, we introduce the Canonical Biplot technique to explore differences in the scientific performance of Spanish CSIC researchers, organised by field (Chemistry and Materials Science) and grouped by academic rank (research fellows and three types of full-time permanent scientists). This method enables us to build a Biplot where the groups of individuals are sorted out by the maximum discriminating power between the different indicators considered. Besides, as confidence intervals are displayed in the plot, statistical differences between groups are liable to be studied simultaneously. Since test hypotheses are sensitive to different sample size effects, sizes for some pairwise comparisons are computed. We have found two gradients: a primary gradient where scientists mainly differ in terms of age, production, number of collaborators, number of highly-cited papers and their position in the byline of the publications; and a second gradient, in which scientists with the same academic rank differ by sort of field.