Survivorship Bias

Looking only at the successes and ignoring the failures, which distorts conclusions. The lecture's example: in WW2 the military wanted to armour the parts of planes with bullet holes, but statistician Abraham Wald said to armour the parts without holes, because planes hit there never returned.

When to use

Use it as a trap to spot when a dataset contains only what survived (for example only the startups that succeeded). Ask what data you do not have, since it matters as much as what you do.

See DS Lec 1.

Appears in

No references yet.