Hybrid Bayesian - frequentist approaches for randomized trial design in small populations

Abstract

Randomized Controlled Trials (RCTs) are considered the gold standard for evaluating medical interventions. In small populations, where resources and patients available for participation in research are scarce, the design and conduct of RCTs is especially challenging. Both main schools of statistical inference (frequentist and Bayesian) have shortcomings in that respect. In this thesis, we suggest methods that combine ideas from both those schools in order to borrow their strengths and mitigate their weaknesses. The focus is in efficient use of prior information (a Bayesian concept) while controlling operational characteristics (a frequentist concern).