Dynamic and control systems often contain uncertain parameters that may result in uncertain predictions. In the interest of quantifying the effects of parameter uncertainties on response variability, this paper develops a stochastic response surface based method for the sensitivity analysis of uncertain parameters. Stochastic response surfaces were firstly constructed to describe the explicit relationships between uncertain parameters and responses. Then partial derivations were performed on the mathematical expressions of stochastic response surfaces in order to obtain sensitivity indices that simultaneously embody the effects of parameter means and standard deviations. Lastly, the developed method has been verified against a numerical cantilever beam containing uncertain geometric and material parameters. The sensitivity analysis results were compared with those given by the analysis of variance method.