Predictive aspects of some biobliometric processes.

Abstract

A statistical model for any bibliometric situation, whether i t be a frequency-of-citation distribution over a population of authors, a frequency-of-circulation distribution for a library collection or a Bradford type analysis of journal relevance in a field, is of most value when it can be used to guide future action. The model can be used as a predictive tool only if a time parameter features in the model explicitly, in which case we are obliged to consider not bibliometric distributions but bibliometric processes. We consider various models for such processes and suggest ways in which they can be used for predictive purposes. In many cases, since it will not be obvious, on the basis of available data, which is the most appropriate model, we need to compare the predictions arising from the various models. Consideration of the moment structure as a means of distinguishing between the models is also discussed