Winner's Curse¶
Definition
AI-generated
Winner's curse is the systematic over- or underestimation of genetic effect sizes for variants selected because they exceed a significance threshold, because conditioning on crossing that threshold truncates the sampling distribution of the estimate.
Topics
Why it matters in GWAS¶
Lead SNP effects used for PRS, follow-up power calculations, or Mendelian randomization can be biased unless selected hits are adjusted or externally replicated.
Example usage¶
"The methods section includes Winner's Curse to support interpretation of the primary findings."
Related terms¶
References¶
- Zhong H, Prentice RL. (2008). Bias-reduced estimators for odds ratios in genome-wide association studies. Biostatistics.
- GWASTutorial: Winner's curse.
Last updated (UTC · Git history)