Stochastic models of base station infrastructure deployment by multiple mobile operators can be an invaluable tool for deriving fundamental results about wireless network sharing. In this paper, we study stochastic geometry models for a shared cellular network consisting of base stations deployed by multiple mobile operators, based on real cellular network data coming from three European countries. Relying on a statistical approach as well as the evaluation of wireless network performance metrics, we show that the log-Gaussian Cox process provides the most compelling fitness results with real multi-operator base station deployment patterns and a model that offers some degree of analytical tractability. The model captures the fact that, in urban areas, there is strong correlation between the locations where the base stations of different operators are deployed. In contrast to that, in rural areas we observe some repulsion between antenna locations of different operators. Moreover, we observe that the behavior which can be modelled with the help of these processes occurs over and over again for similar areas in different countries, which suggests universality of the proposed models.