A sensitivity analysis can be used to examine the extent to which random error in exhaust emission regressions influences the models' results. In this study, the analysis is conducted by first simulating the random error by introducing small errors in the dependent emission variables (Nitrogen Oxide emissions (NOx) and Non-Methane Hydrocarbon emissions (NMHC)) and then by examining the resultant effects on the models' predictability. Previously published data from EPA's testing programs (ATL-Phase I and ATL-Phase II) are used to generate the regression models relating exhaust emissions of NOx and NMHC to fuel parameters oxygen, sulfur, E200, E300, olefins, aromatics and RVP. Two different regression techniques are used to generate the baseline models for the sensitivity analyses. A sensitivity analysis is run for each technique and the results are compared.