The likelihood ratio test for the change point problem for exponentially distributed random variables (Q1119304): Difference between revisions

Let \(x_ 1,x_ 2,...,x_ i,...,x_{n+1}\) be \(n+1\) independent and exponentially distributed random variables with intensity \(\lambda_ 1\) for \(i\leq \tau\) and \(\lambda_ 2\) for \(i>\tau\), where \(\tau\) as well as \(\lambda_ 1\) and \(\lambda_ 2\) are unknown. The authors consider the likelihood ratio (LR) statistic for testing \(H_ 0:\) \(\lambda_ 1=\lambda_ 2\) versus \(H_ 1:\) \(\lambda_ 1\neq \lambda_ 2\), and show that the asymptotic null-distribution of the LR statistic is an extreme value distribution. It is also shown that the test is optimal in the sense of Bahadur although the Pitman efficiency is zero. Furthermore, simulation shows good power for those values of n which are relevant to most applications \((20<n<200)\).