Theory 02/Existing data/Clinical trials
Effects Shrink With Sample Size
The claim
Published effect sizes decline monotonically with registered sample size across a field, because small studies that reach significance are selected draws.
01Mechanism
Selection on significance over-rewards lucky small samples; large samples are closer to the true effect.
02Prediction
Effect-size funnel plots should show monotonic decline, not scatter.
03Refutation
Flat effect sizes across sample sizes would remove the selection account.