Cramer-von Mises statistics for testing normality with censored samples

Some key word8: Censored data; Cram6r-von Mises statistics; Goodness of fit; Normality. Pettitt & Stephens (1976) modified Cramer-von Mises type statistics so that tests of goodness of fit could be made for the simple hypothesis with censored data. In this paper the asymptotic theory for the statistics is developed, when tests of fit are made with unknown parameters. The limiting covariance function of the empirical process is derived when estimators from censored samples are used. The theory is applied to find the asymptotic distributions of Cramer-von Mises statistics when testing for normality, with the mean and variance unknown for single-sided and symmetric censoring. Asymptotic percentage points are tabulated for the Cram6r-von Mises statistic, Wn, the Anderson-Darling statistic, A2, and the Watson statistic, U2; the small-sample distributions are investigated by Monte Carlo methods.