Outlier Detector
Find outliers in a numeric CSV column with IQR or z-score rules.
About this tool
Name a numeric column and pick a method. The IQR rule flags values more than 1.5 interquartile ranges beyond the quartiles. The z-score method flags values at least 3 standard deviations from the mean. You get the row indexes and the values.
How to use
How it works
IQR rule: with Q1 and Q3 as the 25th and 75th percentiles, anything below Q1 − 1.5·IQR or above Q3 + 1.5·IQR is an outlier. Z-score: |x − mean| / sd ≥ 3, using the sample standard deviation. IQR is robust to the outlier itself. Z-score is not, and in small samples it can never reach 3 (with n = 7 the largest possible z is about 2.2).
IQR fences: Q1 − 1.5·IQR and Q3 + 1.5·IQR z = (x − mean) / sd, flag if |z| ≥ 3
Worked example
A value of 30 among small numbers
- Input
- Column y: 2.1, 3.9, 6.2, 7.8, 10.1, 12.2, 30
- Output
- IQR: count 1, rowIndexes [6], values [30] z-score: count 0
- Walkthrough
- The z-score of 30 is about 2.1, below the cut-off of 3, so only the IQR rule catches it. With so few rows a z-score of 3 is nearly unreachable.
Limitations
- The multiplier (1.5) and the z cut-off (3) cannot be changed.
- Single column at a time. Missing values are ignored.
- It only flags rows; it does not remove them.
Questions
How do I find outliers in a CSV?
Which method should I use?
Why did the z-score find nothing?
Are row indexes zero-based?
Features
- IQR and z-score
- Row indexes and values
- Runs in your browser
Related resources
Guides and videos from around the web—opens on YouTube or the publisher’s site.
SQL Tutorial - Full Database Course
freeCodeCamp.org
SQL and tabular data sit next to CSV/JSON data tools.
Watch on YouTubeMachine Learning for Everybody
freeCodeCamp.org
Data prep context before ML-oriented helpers.
Watch on YouTube
- Data Index
Math is Fun
Stats explainers when summarizing columns.
- Model for Tabular Data and Metadata on the Web
W3C
Tabular data model behind CSV-oriented workflows.