Citation

Sparks G, Horosko A, Smith A. J. Transp. Res. Forum 1988; 29(1): 24-27.

Copyright

(Copyright © 1988, Transportation Research Forum)

DOI

unavailable

PMID

unavailable

Abstract

Considerable debate has emerged in Canada in terms of the relative safety of different types of large trucks (i.e., double trailer units versus single trailer units; A-Train versus B-Train versus C-Train configurations, etc.). This paper presents an analysis of the sample size required in order to have a reasonable expectation of detecting statistically significant differences in safety experience for varying levels of actual differences in accident rates. The results indicate that the sample size required to detect a 10-20% difference in (actual) accident rates is in the order of 1-2 x 10 to the 9th power vehicle kms of travel. Smaller samples could only be expected to detect larger actual differences. These results suggest that: (1) it is very important to include an appropriate statistical analysis as part of accident research analysis if differences in the order of 20% are of interest; and (2) the value of much of the research reported on the relative safety experience of various vehicle types may be limited and/or questionable because of limited sample sizes and/or the lack of appropriate statistical analysis.