Models that accurately describe chemical processes are often intricate involving numerous reacting species and reaction steps. For complex reaction mechanisms, output-species concentration profiles can change dramatically based on the set of values chosen for inputs if they are nondeterministic. A systematic uncertainty analysis can provide insight into the level of confidence of model estimates and aid mechanism reduction. Response surface methods and variants, thereof, require much fewer simulations for the adequate estimation of system uncertainty characteristics. This article focuses on reaction rate constant uncertainty using the stochastic response surface method (SRSM), whereby, uncertain outputs are expressed in terms of a polynomial chaos expansion of Hermite polynomials and engenders such useful properties as the mean and valiance and computation of sensitivity information. SRSM determines the uncertainty propagation characteristics very accurately, while using an order-of-magnitude fewer model simulations than traditional Monte Carlo techniques. Since uncertainty in kinetic rate parameters largely affects the reduction of kinetic models, a framework of analysis is also developed for mechanism reduction considering uncertainty using sensitivity information from SRSM to create good initial sets of reactions for the efficient solution of a multi-period optimization problem. Two case studies-an isothermal supercritical wet oxidation process and a nonisothermal H-2/CO/air combustion process-elucidate the application of this framework of analysis to complex kinetic mechanisms and illustrate the possible ease of computational burden associated with mechanism reduction under uncertainty.