Number 13/Life & ecology/Meta-science
The Selection Ratio
The quantity
The ratio of the published effect to the expected replication effect, as a function of the original study's statistical power.
01Why it matters
It converts the vague expectation that replications shrink effects into a forecast that can be scored before the replication is run.
02Expected value
Close to one at high power, falling steeply below power of about 0.5.
03How it would be measured
Compute from original designs alone; then compare to observed replications.
04What would refute it
If shrinkage is unrelated to power, selection is not the mechanism.