A decomposition for the likelihood ratio statistic and the Bartlett correction - a Bayesian argument (Q803686)

A general high order asymptotic expansion for a multidimensional posterior distribution is established. The assumptions do not require any particular dependence structure of the observations. The main result leads to an expansion for the posterior distribution of the likelihood ratio test statistic. As an application, statements concerning the Bartlett correction are verified. Technically, the paper applies to multidimensional Edgeworth expansions.