For the analysis of the response of structural members made of short-fiber-reinforced-composites. in which short fibers are randomly oriented in the matrix with certain directionality, probabilistic computational techniques are used and compared. Mean direction of fibers and standard deviation of the direction of fibers are selected as the random variables which describe the effective material properties of the short fiber composite materials. Effective material properties are obtained by use of the classical lamination theory and Monte-Carlo integration. Monte-Carlo method, PFEM(probabilistic finite element method using perturbation method) and PEPM(point-estimate-for-probabilistic-moment method) are used for the analysis and the results are compared. As the result of the study, a modified PEPM is proposed. In general, PEPM yields reasonably accurate results for the probabilistic response of the composites. However when the effective properties of the composites are obtained by the aforementioned way PFEM yields unacceptably large errors in the standard deviation of the stresses. For the calculation of Streses, the proposed PEPM yields satisfactory results even though the method requires significantly less computations than the orignal PEPM.