A2085
Title: A simulation study of tvwLDA parameter estimators
Authors: Louisa Kontoghiorghes - Kings College London (United Kingdom) [presenting]
Ana Colubi - University of Giessen (Germany)
George Kapetanios - Kings College London (United Kingdom)
Abstract: A new model, the time-varying weighted Latent Dirichlet Allocation (tvwLDA), has been previously proposed as an extension of the standard Latent Dirichlet Allocation (LDA). Unlike LDA, tvwLDA estimates document-topic and topic-term weights using a weighted log-likelihood over time while accommodating temporal variation in the number of topics. To evaluate the performance of the proposed estimation procedure, Monte Carlo simulations are conducted. The results are used to assess whether the means of the parameter estimators approach the corresponding population parameter values as the sample size increases.
