Groupe d’études et de recherche en analyse des décisions

Statistical estimation problems in meta-analysis

Andrew L. Rukhin National Institute of Standards and Technology (NIST), États-Unis

The principal statistical estimation problem in meta-analysis is to obtain a reliable confidence interval for the treatment effect. Several possible approaches and settings are described. In particular a Bayesian model with non-informative priors and the default data-dependent priors is discussed along with relevant optimization issues.