An empirical adjustment to the likelihood ratio statistic

摘要: Consider a model parameterised by scalar parameter of interest ψ and nuisance λ. Inference about may be based on the signed square root likelihood ratio statistic, R. The statistic R is asymptotically distributed according to standard normal distribution, with error O(n -1 /2). To reduce this approximation, several modifications have been proposed such as Barndorff-Nielsen's modified directed * . In paper, an approximation that can calculated numerically for wide range models. This shown agree order O p (n ). results are illustrated examples.