RobustRank

Rankings can change a lot depending on which data you use, such as which years or groups. An exact ranking is often fuzzy and easily breaks. RobustRank groups the entities into tiers that stay the same across most of the data choices (contexts) you pick. Tiers are less detailed than an exact ranking, but you can trust them more.

Data

Ranking function

Tolerance parameters

Minimum robustness

fraction of contexts where the ranking must hold

Borderline entities per tier

fraction of a tier's entities that may cross a tier boundary

Crossing frequency

fraction of contexts an entity may cross a tier boundary

Penalty-free tier size

tiers with at most this many entities are not penalized

Context specification

Robust ranking

↑ crosses into the tier above↓ crosses into the tier below↕ crosses both ways