Weak convergence of time-sequential censored rank statistics with applications to sequential testing in clinical trials (Q1178947)

The asymptotic behaviour for time-sequential censored rank statistics \(S_ n(t)\) of general form considered by \textit{A.A. Tsiatis} [ J. Am. Stat. Assoc. 77, 855 -- 861 (1982; Zbl 0511.62045] in the two-sample problem of comparing time to failure between two treatment groups \(X\) and \(Y\) is investigated. Let \(F\) and \(G\) denote the distribution of \(X_ i\) and \(Y_ j\), respectively. The result follows by a general weak convergence theorem for a random field constructed on the base of i.i.d. non-negative random variables. To this end, certain maximal inequalities for two parameter processes are developed. The weak convergence of \(\{n^{-1/2}S_ n(t)\}\) in \(D[0,t^*]\) to a Gaussian process under the null hypotheses \(F=G\) and under local alternatives is established. The weak convergence of \(\{n^{-1/2}(S_ n(t)-\mu_ n(t))\}\) under general alternatives \((F,G)\) is also considered, where \(\mu_ n(t)\) is the drift function of \(S_ n(t)\). Applications of the theory to time-sequential tests based on these censored rank statistics are also discussed.