Stopped sum models and proposed variants for citation data

Abstract

It is important to identify the most appropriate statistical model for citation data in order to maximise the potential of future analyses as well as to shed light on the processes that may drive citations. This article assesses stopped sum models and some variants and compares them with two previously used models, the discretised lognormal and negative binomial, using the Akaike Information Criterion (AIC). Based upon data from 20 Scopus categories, some of the stopped sum variant models had lower AIC values than the discretised lognormal models, which were otherwise the best (with respect to AIC). However, very large standard errors were returned for some of these variant models, indicating the imprecision of the estimates and the impracticality of the approach. Hence, although the stopped sum variant models show some promise for citation analysis, they are only recommended when they fit better than the alternatives and have manageable standard errors. Nevertheless, their good fit to citation data gives evidence that two different, but related, processes may drive citations.