Telling the whole story: finding structures in bibliometric information using PCA

Keith Julian, John Rigby

Research output: Other contribution

137 Downloads (Pure)


As bibliometric data is multidimensional, its study, by means of index numbers, especially index numbers in the form of ratios, rarely captures all the information. Across the bibliometric and evaluation literatures there is increasing scepticism about the value that single index numbers give to the understanding of scientific behaviour and its consequences including impact. The authors propose and demonstrate the use of a multivariate approach – principal component analysis - that gives greater insight into the aspects of scientific publication. Principal component analysis is put forward as the most suitable multivariate method available as it does not emphasise any variable and identifies important signals in the data that may not be observed with univariate methods. A data set analysed in a previous piece of work on double-dipping is used here and is subject to PCA which reveals three components, an input related component, an output related component and an outcome (impact) related one. Importantly, citation is shown to be of limited significance in explaining overall variability within the data set.
Original languageEnglish
TypeWorking Paper
Number of pages17
Publication statusPublished - 2017


Dive into the research topics of 'Telling the whole story: finding structures in bibliometric information using PCA'. Together they form a unique fingerprint.

Cite this