Broadly speaking, robust statistics is the study of how to perform reliable statistical inference under distributions that are misbehaved in some way.

“Misbheaving” usually refers to three possible things: