Co-Authors: Daniel E. Acuna
Abstract: Most Published Research Findings Maybe False, But Some Still Worth Continuing
Research has the transformative potential to influence health and technology. However, researchers have shown that under a broad set of reasonable statistical assumptions, most published research findings are likely false (Ioannidis, 2015). Further, recent replication studies have supported this conclusion (e.g., Open Science Collaboration (2015), Camerer et al. (2018)). However, there is another dimension barely discussed: costs vs. benefits. We hypothesize that it is still worth investigating even if most published articles are false. We develop a cost-benefit analysis of research within a Bayesian decision-making framework and apply it to several real-world cases including brain, breast, liver, and lung cancers. Our analysis reveals a cost/benefit ratio to evaluate research decisions and suggests that brain cancer research should be reconsidered further in light of the benefit of the other types of cancer. We discuss the implications of this framework on health and science policies.