Application of Statistical Concepts to the Knock-Rating Problem 380169
CONCLUSIVE evidence that any one make of car has a real, as distinguished from accidental, tendency to rate fuels higher or lower than the average for all cars is difficult to find in existing data, the authors report.
The growing economic and technical importance of antiknock quality, they point out, is making it increasingly desirable to estimate the random distribution of errors that occur in making measurements of knock, either in the laboratory or in road tests. Statistical analysis, they explain, offers a means of appraising the probability of occurrence of errors of various magnitudes. To make a beginning in such a statistical analysis of the available data on fuel ratings is the purpose of their paper, they announce.
The variability, or random error, among successive measurements in road tests was from three to six times higher than the variability among measurements of octane number in the A.S.T.M.-C.F.R. single-cylinder test engine, the paper reports. Also, the variability among ratings made in different makes of cars was only slightly greater than the variability among repeated tests in a single car of one make.
The high variability in road ratings also makes it necessary to obtain a relatively large number of road ratings on any particular gasoline before a reliable comparison can be made between it and other gasolines with the same degree of precision that is obtained in the mean of three or four laboratory ratings on each fuel, the paper concludes.