This research focuses on combining Internet clickstream data from multiple online retailers and develops a stochastic model of cross-site visit-timing behavior to understand how to leverage information from one site to help explain consumer behavior at another. To this end, we incorporate two sources of association in browsing patterns: one for the observable outcomes (i.e., arrival times) of two timing processes and the other for the latent visit propensities across a set of competing sites. The proposed multivariate timing mixture model can be viewed as a generalization of the univariate exponential-gamma model. On the basis of Internet clickstream data collected by Media Metrix, we demonstrate that this general (yet parsimonious) model offers significantly superior performance compared to the naive independent model. This research shows that a failure to account for both sources of association patterns not only leads to poor fit and forecasts, but also generates systematically biased parameter estimates. While the model assuming complete independence across websites is overly reliant on the information on consumer visits within a given site, the proposed approach argues that it is critical to consider consumer browsing behavior across online stores as well as within a given online store. Furthermore, this thesis sheds light on the zero-class (i.e., nonvisitors) problem by combining databases across multiple online retailers.
