B0867
Title: Confidence intervals construction in complex parametric models facilitated by inconsistent estimators
Authors: Samuel Orso - University of Geneva (Switzerland) [presenting]
Mucyo Karemera - University of Geneva (Switzerland)
Stephane Guerrier - University of Geneva (Switzerland)
Maria-Pia Victoria-Feser - University of Geneva (Switzerland)
Abstract: A novel approach is presented to constructing confidence intervals within complex parametric models where obtaining a consistent estimator is not readily feasible. Existing methods have been developed to derive a consistent point estimator from an initially "easy-to-obtain" but inconsistent estimator. However, constructing confidence intervals from these point estimators using conventional approaches (such as asymptotic normality or bootstrap) poses significant computational challenges. In the proposed method, a distribution for the parameters of interest is directly derived from the inconsistent estimators, bypassing the need for a consistent point estimator. The approach offers a computational shortcut and serves as an alternative to bootstrap methods, presenting its own advantages. It is demonstrated, under general conditions, the first-order accuracy of the percentile confidence intervals constructed using the distribution obtained from the new method. Furthermore, simulation studies are conducted to illustrate these findings.