This paper makes three contributions: (1) the paper provides a better understanding of online behavior by showing the main drivers of Internet portal choice, (2) the rich data allow for a deeper understanding of brand substitution patterns than previously possible and (3) the paper introduces a wider statistics community to a new data opportunity and a recently developed method.
"Household-Specific Regressions Using Clickstream Data." Statist. Sci. 21 (2) 247 - 255, May 2006. https://doi.org/10.1214/088342306000000150