Outlier tests
When is a sample really an outlier? Two levels, two tests: one value in one feature, or one whole sample across many. Then the part that matters most, which is what to do after the flag.
In this lesson
- The Grubbs test for a single value.
- The Mahalanobis distance for a whole sample.
- Two quick rules you can apply by eye.
- Check the sample sheet, QC metrics and batch before removing anything.
Narration is an AI-generated voice. Data shown are simulated. The video plays from YouTube in privacy-enhanced mode and loads only when you press play.