Direct comparison of bibliometric models

摘要

This study describes a technique for statistically comparing bibliometric models, and illustrates its use with three different examples. The technique is based on the idea of comparing full and restricted models as developed in analysis of variance, regression, and log-linear models. In bibliometrics, any two models where one is a special case of the other can be thought of as a full model and a restricted model. One can use the likelihood-ratio chi-square statistic, which has gained acceptance with log-linear models, as a test statistic to directly compare the full model and the restricted model. The first two examples involved Lotka's law. In the first example we investigated the feasibility of applying a single set of global parameter values to eight different author productivity distributions drawn from two different disciplines. In the second example we looked at whether or not a finite maximum productivity level was necessary as an additional parameter in Lotka-type models of author productivity. The final example compared three different forms of a model of library circulation frequencies.