A Saddlepoint Approximation to the Distribution Function of the Anderson-Darling Test Statistic

The Anderson-Darling goodness-of-fit test has a highly skewed and non-standard limit distribution. Various attempts have been made to tabulate the associated critical points, using both theoretical approximations and simulation methods. We show that a standard saddlepoint approximation performs well in both tails of the distribution. It is markedly superior to other theoretical approximations in the lower tail of the distribution. Author Contact: Professor David Giles, Department of Economics, University of Victoria, P.O. Box 1700 STN CSC, Victoria, B.C., Canada, V8W 2Y2 e-mail: [email protected]; FAX: (250) 721 6214; Phone: (250) 721 8540