Bibliometrics, as a discipline, is as hard as it gets, since it's trying to measure, or even, predict, an outcome that perhaps only years or decades can validate.
Among the statistical models used, complex as they may be, I find missing the details of how the enchilada is made. As in, when we publish a paper, we are often told by the editor, "you are 5 pages over your 5 page limit", and, "you are 100 citations over your 50 citation limit". So we rewrite the manuscripts (hence preprints are often better, at least in the honesty and clarity of the citations) to compress both the text and the references, favouring reviews or simply skipping those that may be considered common knowledge or which "merely" confirm prior claims. Now try to model that. I hope there's focus on preprints for more proper studies of attribution and discovery chains.
#academia #ScientificPublishing