Benchmarking research performance at the University. Level with information theoretic measures

Abstract

This paper presents a new method for comparing universities based on information theoretic measures. The research output of each academic institution is represented statistically by an impact-factor histogram. To this aim, for each academic institution we compute the probability of occurrence of a publication with impact factor in different intervals. Assuming the probabilities associated with a pair of academic institutions our objective is to measure the Information Gain between them. To do so, we develop an axiomatic characterization of relative information for predicting institution-institution dissimilarity. We use the Spanish university system as our scenario to test the proposed methodology for benchmarking three universities with the rest as a case study. For each case we use different scientific fields such as Information and Communication Technologies, Medicine & Pharmacy, and Economics & Business as we believe comparisons must take into account their disciplinary context. Finally we validate the Information Gain values obtained for each case with previous studies.