Asymptotic properties for the sequential CUSUM procedure (Q1113593)

Let \(X_ 1,X_ 2,..\). be independent random variables. Consider the changepoint problem \(H_ 0:\) \(X_ 1,X_ 2,..\). are identically distributed versus \(H_ A:\) exactly one changepoint occurs. Define \(\xi_ i=F_{i-1}(X_ i)\), where \(F_ i\) is the empirical distribution function of the \(X_ j\). Assume \(H_ 0\) holds. Under appropriate conditions it is shown that for the one-sided stopping rule \[ \tau (m)=\inf \{k:\quad k\geq m,\quad \sqrt{12}\sum^{k}_{i=2}(\xi_ i-1/2)\geq \sqrt{m} g(k/m)\}: \] \[ \lim_{m\to \infty}P\{\tau (m)<\infty \}=P\{\sup_{1\leq t<\infty}W(t)/g(t)>1\}. \] Results for two-sided stopping rules and asymptotic results for the alternative hypotheses are also given.