A large deviation result for the likelihood ratio statistic in exponential families (Q801418)

In this paper we consider exponential families of distributions and obtain under certain conditions a uniform large deviation result about the tail probability \(P_{\delta}(\phi_{\delta}(\bar X_ n)>\epsilon)\), \(\epsilon >0\), where \(\delta\) is the natural parameter and \(\phi_{\delta}(\bar X_ n)\) is the log likelihood ratio statistic for testing the null hypothesis \(\{\) \(\delta\) \(\}\). The technique involves approximating certain convex compact sets in \(R^ k\) by polytopes, then estimating the probability contents of associated closed half-spaces, and counting the number of these half-spaces. Some examples are given, among them the multivariate normal distribution with unknown mean vector and covariance matrix.