The moisture susceptibility of asphaltic concrete mixtures can be evaluated using a one-tail lower t-test for the mean difference between dry and wet tensile strengths. The practical number of specimens to be used for moisture susceptibility testing in the laboratory should also be valid statistically. On the basis of the data from 960 laboratory specimens, the reliabilities for using various numbers of samples for the moisture susceptibility test were evaluated by several statistical analyses. The power of the t-tests, a measure of the reliability of the moisture susceptibility tests, was also evaluated on the basis of sample size, difference between dry and wet means, and variances among samples. The analyses showed that the decision derived from the t-test based on small sample sizes had low reliability. Factors that could possibly influence the power of the test were examined. It was found that the most likely method for improving reliability was to increase the number of samples used. Therefore, it is recommended that the minimum number of samples be determined based on sample variance, difference between dry and wet means, and the expected importance of the result. This paper appears in transportation research record no. 1228, Asphalt mixtures and asphalt chemistry.
Abstract