Saddlepoint Approximation near the Endpoints of the Support


Rungao Jin and John Robinson


Research Report 99-1
Date: 12 January 1999


For some cases, non-normal asymptotic distributions of tilted means are used to give accurate saddlepoint approximations at the extremes of the support where asymptotic normality of the tilted means fails. Also new relative error rates are obtained near the endpoints.

Key phrases

saddlepoint methods. tilting. binomial statistic. hypergeometric statistic. Wilcoxon statistic.

AMS Subject Classification (1991)

Secondary: 62G20


The paper is available in the following forms:
TeX dvi format:
1999-1.dvi.gz (36kB) or 1999-1.dvi (112kB)

PostScript: (81kB) or (269kB)

To minimize network load, please choose the smaller gzipped .gz form if and only if your browser client supports it.

Sydney Mathematics and Statistics