A1594
Title: Latent space directed counting network models and their application to citation networks of statistical journals
Authors: Ji Zhu - University of Michigan (United States) [presenting]
Abstract: Impact factors evaluate a journal's significance. Specifically, for a given year, it is defined as the ratio between the number of citations received this year by publications in this journal in two preceding years and the number of publications in this journal in two preceding years. Although popular and straightforward, impact factors have two limitations. First, they do not distinguish between citations from different domains, counting all citations for a journal equally. Second, impact factors are one-dimensional metrics that overlook mutual citation information between journals, a two-dimensional data resource, which could provide valuable insights. A latent space directed counting network model is introduced, that explores latent variables driving the citation pattern among journals in the same domain through the analysis of mutual citation counts. The proposed model simultaneously takes into account a journal's general significance and its significance through interactions with other journals in the domain. Likelihood-based estimators of the parameters with their statistical optimality established are introduced. A simulation study verifies the theory and shows the effectiveness of the algorithm. A dataset consisting of mutual citation information among 104 statistical journals is collected and cleaned. Fitting the data with the latent space directed counting network model, meaningful and interpretable findings are uncovered that are not conveyed by impact factors.
