Caching of video files at the wireless edge, i.e., at the base stations or on user devices, is a key method for improving wireless video delivery. While global popularity distributions of video content have been investigated in the past, and used in a variety of caching algorithms, this paper investigates the statistical modeling of the individual user preferences. With individual preferences being represented by probabilities, we identify their critical features and parameters and propose a novel modeling framework as well as a parameterization of the framework based on an extensive real-world data set. Besides, an implementation recipe for generating practical individual preference probabilities is proposed. By comparing with the underlying real data, we show that the proposed models and generation approach can effectively characterize individual preferences of users for video content.