Gaussian approximation of signed rank statistics process (Q1065473): Difference between revisions

Let \(Y_{Ni}\), \(1\leq i\leq N\), \(N\geq 1\), be independent random variables with respective distribution functions \(F(y-\Delta d_{Ni})\) where \(d_{Ni}\) are real numbers and \(\Delta\) is an unknown real parameter in [0,1]. Consider the signed linear rank statistic \[ S_{\Delta N}=\sum^{N}_{i=1}C_{Ni}\phi ([R^{\Delta}_{Ni}/(N+\quad 1)])Sgn Y_{Ni}. \] The authors show that the process \(\{S_{\Delta N}- S_{0N}-ES_{\Delta N}\), \(0\leq \Delta \leq 1\}\), properly normalized, converges weakly to a Gaussian process. As an application, the asymptotic distribution of the properly normalized length of a confidence interval for \(\Delta\) is derived.