A Small Sample Power Study of the Anderson-Darling Statistic and a Comparison with the Kolmogorov and the Cramer-Von Mises Statistics
Abstract
The Anderson-Darling goodness-of-fit procedure emphasizes agreement between the data and the hypothesized distribution in the extremes or tails. An improved table of the quantiles of the Anderson-Darling statistic, useful for small sample sizes, was constructed using the Cray-2 supercomputer. The power of the Anderson-Darling test is compared to the Kolmogorov and the Cramer-von Mises tests when the null hypothesis is the normal distribution and the alternative distributions are the Cauchy, the double exponential, and the extreme value distributions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1989
- Accession Number
- ADA215168
Entities
People
- Henry B. Tingey
- Linda L. Moss
- Malcolm S. Taylor
Organizations
- Ballistic Research Laboratory