Predicting publication productivity for researchers: A latent variable model

Abstract: This study provided a model for the publication dynamics of researchers, which is based on the relationship between the publication productivity of researchers and two covariates: time and historical publication quantity. The relationship allows to estimate the latent variable the publication creativity of researchers. The variable is applied to the prediction of publication productivity for researchers. The statistical significance of the relationship is validated by the high quality dblp dataset. The effectiveness of the model is testified on the dataset by the fine fittings on the quantitative distribution of researchers' publications, the evolutionary trend of their publication productivity, and the occurrence of publication events. Due to its nature of regression, the model has the potential to be extended for assessing the confidence level of prediction results, and thus has clear applicability to empirical research.