This paper uses a mixed-effects analysis of covariance model (with both fixed and random effects) to characterize mileage-dependent emissions profiles for any given group of vehicles having a common model design. Such profiles are useful for evaluating, for example, how emissions will change over time within a new line of vehicles. The U.S. Environmental Protection Agency uses these types of evaluations to certify whether or not new models conform to existing emissions standards. Given such a group of vehicles, the statistical model introduced in this paper describes both the average emissions profile for that group while also accounting for individual vehicle variability among vehicles within the group. The model can be used to provide realistic confidence bounds for the average emissions deterioration profile within a given group, therefore allowing accurate emissions comparisons of multiple groups. The approach is illustrated with a sample of emissions data from two types of vehicles: natural gas Dodge Ram vans and gasoline Dodge Ram vans (all from the 1992–94 model years). The population profile for nonmethane hydrocarbons is explored. The results indicate the presence of vehicle-to-vehicle variation within each vehicle type. This variation leads to confidence profiles that can be markedly different (but more appropriate) than what would be obtained from a simple fixed-effects regression model. The results highlight the potential for incorrectly characterizing emissions profiles whenever decisionmakers rely on standard regression techniques.

Publication Date



ISSN:1094-8848 Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works in February 2014.

Document Type


Department, Program, or Center

Sustainability (GIS)


RIT – Main Campus