In mass production of RF/microwave circuits, manufacturing tolerances and process variations will make the circuits coming out of the same production line to behave differently. Statistical modeling of RF/microwave passive and active devices taking into account these random variations are important for yield-driven design. While statistical modeling for small-signal linearized device models is fairly mature, the statistical large-signal model for nonlinear devices remains a challenge, due to the expense of nonlinear measurements, the difficulties in nonlinear statistical problems, and the cost of EM evaluations of packaging structures. We describe recent progress in this area exploiting statistical space mapping concepts. Using such a concept, we expand a large-signal nominal model into a large-signal statistical model. The nominal model is extracted or trained from one complete set of large-signal data. The statistical property of the model is achieved by a dynamic mapping between the nominal model and the statistical samples of a given population of devices. This method reduces the otherwise prohibitive task of creating a population of nonlinear device models into a simplified one of creating a population of mapping functions. Two mapping approaches are presented, one use linear mapping for small-variations in parameters, and another using neuro-space mapping for large variations in device parameters (L. Zhang, Q.J. Zhang and J. Wood, “Statistical neuro-space mapping technique for large-signal modeling of nonlinear devices,” IEEE Trans. Microwave Theory Tech., vol. 56, no. 11, pp. 2453–2467, Nov. 2008.). Furthermore, yield-based design of microwave circuits, taking into account random variations in circuit parameters will also be described. Examples of statistical modeling of RF/microwave transistors, and use of the models in statistical analyses of microwave passive and active circuits will also be presented.