Robust small sample properties for likelihood based confidence regions

Many measurement applications, in particular for biological, biomedical and biochemical systems, deals with the problem that only short data records are available. This problem implies that many powerful statistical tools are not applicable since these tools are based on asymptotic results. No user-friendly guidelines are available when dealing which such sparse datasets.In contrast to an estimator, one often uses an interval estimate when dealing with small data-records. An interval estimate specifies a range within which the parameter is estimated to lie. Confidence intervals are commonly reported in tables or graphs along with point estimates of the same parameters, to show the reliability of the estimates. In this paper, we study the robustness of asymptotically ´optimal´ confidence regions to small/finite sample effects. Our study reveals immediate guidelines to the user when dealing with short records.