SUMMARYA class of random composite materials with statistically inhomogeneous microstructure, for example, functionally graded materials is considered in this paper. The microstructures inside a component are gradually varying in the statistical sense. In view of this particularity, a novel statistical second-order two-scale (SSOTS) method is presented to predict the mechanical properties, including stiffness, and elastic limit. To develop this method, the microstructures of statistically homogeneous, and inhomogeneous materials are represented. In addition the SSOTS formulas are derived based on normalized cell depending on the position variables by a constructing way, and the algorithm procedure is described. The mechanical properties of the different inhomogeneous materials are evaluated. The numerical results are compared with the experimental findings. It shows that the SSTOS method is effective.