Bayesian estimation of manufacturing effects in a fuel economy model.

The analysis of fuel economy data results in estimates of the technology utilization by manufacturer and vehicle line. The analysis employs a hierarchical Bayesian regression model with random components representing vehicle lines and manufacturers. The model includes predictor variables which describe vehicle features, such as type of transmission, and vehicle line specific measurements, such as compression ratio. Non-informative priors with novel modifications are used and the Bayes estimates are obtained by use of Gibbs sampling. The results show there is substantial variability among manufacturers in efficiently utilizing technology for fuel economy. [ABSTRACT FROM AUTHOR]