A test for the presence of stochastic ordering under censoring: the \(k\)-sample case (Q2304252)

Let \(X_1\), \(X_2\) be two nonnegative random variables with distribution functions \(F_1\), \(F_2\) and survival functions \(\overline{F}_1=1-F_1\), \(\overline{F}_2=1-F_2\), respectively. Distribution function \(F_1\) is said to be stochastically smaller than \(F_2\), denoted by \(F_1\preceq F_2\), if \(\overline{F}_1(x)\leqslant \overline{F}_2(x)\) for all real \(x\). The author of the paper presents the new empirical likelihood-based test for hypothesis \(\mathcal{H}_0\): \(F_1=F_2=\cdots=F_k\) against \(\mathcal{H}_1-\mathcal{H}_0\), where \(\mathcal{H}_1\) is the stochastic ordering alternative given by \(\mathcal{H}_1\): \(F_1\preceq F_2\preceq\cdots\preceq F_k\) and \(\{F_1, F_2,\dots,F_k\}\) is a collection of \(k\) continuous distribution functions.