Informative Bayesian Modeling With Applications to Media Data

This dissertation consists of three main parts. Each part develops an application or methodology within the Bayesian framework. The first is a study of multi-channel media consumption patterns for US audiences during the 2010 FIFA World Cup using a Bayesian data fusion strategy. We utilize the aggregated television ratings in the estimation, to incorporate additional data that is on a different scale than the individual-level on alternative media platforms. The second study proposes an information integration method, called the information reweighted prior (IRP) approach, to incorporate external information via prior distributions through reweighting. We demonstrate the effectiveness of IRP with both simulated and real panel choice datasets, and show that `sensible' external information, even if with considerable uncertainty, can improve inferences for quantities of interest. The third study proposes a rank enhanced likelihood (REL) approach to utilize ranking information via re-construction of the likelihood. We demonstrate the effectiveness of REL with simulated datasets, and show that utilizing REL can also improve posterior inferences.