Dilemma: Outliers

Back to Upright overview

When screening my data, I find that there is one extreme observation. What do I do?

A: Nothing, it is part of my theoretical sample for a reason.
B: I look for information on the observation, trying to find qualitative reasons for the deviance. If there is a good explanation for its position, it must be a niche observation. Since it is a part of my theoretical sample, I leave it in. However, if there is no explanation for its position, I leave it out as it is there either due to measurement error or response bias.
C: I look for information on the observation, trying to find qualitative reasons for the deviance. If there is a good explanation for its position, it must be a niche observation. Despite my theoretical sampling, there is no place for such anomalous observations, so I leave it out. If there is no explanation for its position, I leave it in to avoid potential sampling bias.
D: I let a colleague review the data and follow her advice whatever it is.

(We gratefully acknowledge permission to reproduce this dilemma from the “Dilemma Game. Professionalism and Integrity in Research” as developed by the Erasmus University Rotterdam (link to the game) in this tool.)