Statistical Significance and Effect Sizes of Differences among Research Universities at the Level of Nations and Worldwide based on the Leiden Rankings

TL;DR

This study uses statistical analysis and network tools on Leiden Rankings data to classify research universities by significance and effect size, revealing limited meaningful distinctions beyond broad groups.

Contribution

It introduces a combined approach using stability intervals, significance tests, and effect sizes to classify universities, highlighting the limitations of such groupings.

Findings

Groupings are consistent across significance levels but uncorrelated with effect size-based classifications.

Effect sizes between universities are generally small (w < 0.2).

Distinctions beyond three or four university groups may lack meaningfulness.

Abstract

The Leiden Rankings can be used for grouping research universities by considering universities which are not statistically significantly different as homogeneous sets. The groups and intergroup relations can be analyzed and visualized using tools from network analysis. Using the so-called "excellence indicator" PPtop-10%--the proportion of the top-10% most-highly-cited papers assigned to a university--we pursue a classification using (i) overlapping stability intervals, (ii) statistical-significance tests, and (iii) effect sizes of differences among 902 universities in 54 countries; we focus on the UK, Germany, Brazil, and the USA as national examples. Although the groupings remain largely the same using different statistical significance levels or overlapping stability intervals, these classifications are uncorrelated with those based on effect sizes. Effect sizes for the differences between universities are small (w <.2). The more detailed analysis of universities at the country level suggests that distinctions beyond three or perhaps four groups of universities (high, middle, low) may not be meaningful. Given similar institutional incentives, isomorphism within each eco-system of universities should not be underestimated. Our results suggest that networks based on overlapping stability intervals can provide a first impression of the relevant groupings among universities. However, the clusters are not well-defined divisions between groups of universities.