Statistical Evaluation of Different Sample Sizes for Local Calibration Process in the Highway Safety Manual

Abstract

An approach to determine a statistically reliable sample size for developing local calibration factors (LCFs) was proposed to complement the Highway Safety Manual's (HSM) sampling guidance. The HSM suggests a minimum sample size of 30 to 50 sites per facility type with at least 100 annual crashes. However, the HSM fails to provide clear guidance on how to determine a minimum sample size to ensure the statistical reliability of LCFs. The proposed approach based on the finite population correction (FPC) factor determined minimum sample sizes by considering trade-offs between the desired error levels of the estimated LCFs, confidence levels, and sample standard deviations. The sample sizes by facility types were drawn on the basis of various statistical assumptions; then they were assured by the comparisons between FPC-based samples and the HSM-based samples. LCF values estimated from the HSM-based sample sizes yielded inconsistent reliabilities depending on the facility types. In contrast, those estimated from the samples by the FPC-based approach satisfied the desired reliabilities of the LCFs for all facility types.