A combined loglinear/MDS model for mapping journals by citation analysis

Loglinear analysis of interjournal citations permits objective evaluation of the network of journals in and around a discipline. Citation frequency is modeled as a probabilistic Poisson process, with its expectation being the product of the “influence” of the cited journal on the citing journal, and of the citing journal’s “receptivity.” The “influence” is itself modeled as the product of the cited journal’s “importance,” and of the “similarity” between the two journals. Loglinear analysis is used to extract maximum likelihood estimates of journals’ importances and receptivitles and a matrix of similarities from their citation matrix. Multidimensional scaling of the derived similarities matrix provides an interpretable map of the journais’ relative configuration. As an application of the method, the major marketing journals are mapped relative to journals of related disciplines, showing the influence that the psychology and management disciplines have had upon the marketing literature.