In this paper, we study a novel transmission framework based on statistical channel state information (SCSI) by incorporating edge caching and beamforming in a fog radio access network (F‐RAN) architecture. By optimizing the statistical beamforming and edge caching, we formulate a comprehensive nonconvex optimization problem to minimize the backhaul cost subject to the BS transmission power, limited caching capacity, and quality‐of‐service (QoS) constraints. By approximating the problem using the l₀‐norm, Taylor series expansion, and other processing techniques, we provide a tailored second‐order cone programming (SOCP) algorithm for the unicast transmission scenario and a successive linear approximation (SLA) algorithm for the joint unicast and multicast transmission scenario. This is the first attempt at the joint design of statistical beamforming and edge caching based on SCSI under the F‐RAN architecture.