2005
Why Most Published Research Findings Are False
The arithmetic showing most published discoveries are probably wrong — and why the replication crisis was inevitable.
“Simulations show that for most study designs and settings, it is more likely for a research claim to be false than true.”
The base rate problem
Ioannidis models science as a signal-detection problem. When true effects are rare, studies are underpowered, and many hypotheses get tested, false positives dominate published literature.
What drives false claims
Small sample sizes, flexible analysis, financial incentives, and publication bias each inflate the false discovery rate. The math is general; the sting is specific to biomedicine and anywhere p-values gate careers.
A paper people cite and ignore
Fifteen years later, replication crises confirmed the shape of the argument. The paper belongs in any archive about how knowledge is produced, not just how models are trained.
- collection
Why big systems behave in ways none of their parts intend — and how they break.
← previous · 2005
Power Laws, Pareto Distributions and Zipf's Law
next · 2008 →
Bitcoin