A uniform bound on the characteristic function of the exponentially tilted null-distribution of a simple linear rank statistic and its rate of convergence (Q5950021)

For properly standardized simple linear rank statistics based on independent continuous random variables, \textit{W. R. van Zwet} [in: Nonparametric statistical inference. Vol. II. Colloq. Math. Soc. János Bolyai 32, 889-909 (1982; Zbl 0514.62027)] obtained an upper bound for the characteristic function of these simple linear rank statistics. The present article considers the special case of the null-hypothesis, and obtains a generalization of the van Zwet's upper bound for the characteristic function of the exponentially tilted distribution. This result solves an essential step in the procedure for obtaining a rate of convergence for saddle-point expansions for more general rank statistics. In particular, the new upper bound is used within an authors' previous paper [Can. J. Stat. 28, No. 1, 137-149 (2000)] when proving the rate of convergence for their saddle-point expansion of the Wilcoxon-Mann-Whitney null-distribution. Examples satisfying the conditions under which this result was obtained are discussed.