Reproducible Research and Correct Conclusions

• Activity Number: 353
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Naomi Altman*+
• Companies: Penn State
• Keywords: p-value ; false discovery rate ; false nondiscovery rate ; PPV ; positive predictive value ; negative predictive value.

Abstract

The demonstration that even highly influential research may not be reproducible and the reasons for irreproducibility have recently become topics of hot debate. Ioannidis (2005a) used a search of highly cited papers in highly respected medical journals to bring attention to the fact that often the results were contradicted or weakened in subsequent studies. In a provocative follow-up commentary, Ioannidis (2005b) used probabilistic arguments to demonstrate that {\em most} published research findings are likely to be false. Since then, the issue has been taken up in both the research literature and the popular press shedding much heat and some light. This article continues the discussion, defining positive and negative reproducibility and negative predictive power.

p-value; false discovery rate; false non-discovery rate; PPV; positive predictive value; negative predictive value.