This paper investigates global sensitivity analysis as applied to reaction mechanisms. It uses the HDMR (high-dimensional model representation) expansion and the features of the sensitivity indices to, explore the probability density function (pdf) of predicted target outputs that results from the uncertainties in the rate coefficients. We study the auto ignition of H-2/O-2 mixtures, where the pdf describes the uncertainty in ignition delay times due to the uncertainties in the rate coefficients. The global sensitivity analysis in conjunction with the HDMR expansion allows for the deconstruction of the pdf in several different ways. These deconstructions allow the features of the pdf to be understood in terms of the constitute reactions in much finer detail than the study of a variance decomposition alone.