An uncertainty analysis method is proposed here, which uses Fourier Amplitude Sensitivity Test (FAST) and Stepwise Regression Technique (SRT). This method is a compromise between the approximation method (response surface method (RSM) or moments method) and Monte Carlo method (MCM). It is concluded that: 1. FAST gives the partial variance for each input parameter, which can be used as global sensitivity ranking between input parameters, with moderate sampling point compared to crude MCM. 2. SRT is a good tool to construct the later-used first- or second-order response surface model consisting of comparatively important parameters. 3. The combined uncertainty analysis method using FAST and SRT can be used for uncertainty/sensitivity analysis of the large computer codes with moderate cost and it will be a useful tool to analyze the feasibility of the newly developed, highly uncertain system models.