A New Statistical Model for Predicting Customer Octane Satisfaction Using Trained-Rater Observations 801356
Customer-determined and rater-determined octane number requirements have historically-been compared by the Coordinating Research Council (CRC) using the N+1 and Quantified approaches. However, these analysis techniques oversimplify the relationship between customer and rater octane number requirements because they do not consider the bivariate and nonnormal nature of the data.
Using a new statistical technique designed specifically for the analysis of bivariate nonnormal data, a model of the relationship between customer and rater octane number requirements has been developed. The model utilizes data from the 1978 CRC Customer/Rater Octane Number Requirement Program and was verified by “prediction” of the results from a similar CRC program conducted in 1975. Based on rater data from the 1979 CRC Octane Number Requirement Survey, the model predicts that 91 percent of the customers driving 1979 model year cars are satisfied with 91 Research octane number (RON) unleaded gasoline.