The signed log likelihood ratio appears in testing hypothesis on the mean against a one-sided alternative. In this article we shall show that this statistic is a good normalizing transformation. It is shown that the density of the signed log likelihood ratio has an absolutely convergent Gram-Charier expansion. This theorem, based on the saddlepoint technique, is valid for an exponential family under certain conditions.
All Science Journal Classification (ASJC) codes
- Statistics and Probability