Outliers: Keep Them, Remove Them, or Investigate?

A single weird value can distort your entire analysis. But outliers are also where the most interesting stories live. Knowing when to remove them is an art.

30 min · Reviewed 2026

The Value That Does Not Fit

You are analyzing student test scores. 98 percent of scores are between 60 and 100. One score is 3,400,000. Obviously a data-entry error. Remove it. But sometimes outliers are the real story: the one billionaire in an income dataset, the one fraudulent transaction, the one patient whose recovery changed medicine.

Three kinds of outliers

Type	Example	What to do
Error	Height of 2,340 cm	Remove or fix
Extreme but valid	CEO earning $500M	Keep, but note it
Anomaly of interest	Fraudulent transaction	That IS the signal

Detection methods

IQR rule: flag anything below Q1 - 1.5*IQR or above Q3 + 1.5*IQR
Z-score: flag anything with |z| > 3
Isolation Forest: ML-based anomaly detection, works in high dimensions
Domain knowledge: the best detector, often ignored

import pandas as pd
import numpy as np

df = pd.read_csv('data.csv')

# IQR method
Q1 = df['value'].quantile(0.25)
Q3 = df['value'].quantile(0.75)
IQR = Q3 - Q1
low = Q1 - 1.5 * IQR
high = Q3 + 1.5 * IQR

outliers = df[(df['value'] < low) | (df['value'] > high)]
print(f'Found {len(outliers)} outliers out of {len(df)} rows')

# Inspect before removing
print(outliers.head(20))IQR-based outlier detection

Decision framework

Plot your data first — outliers often jump out visually
Check whether the outlier is likely an error (units, typos)
If it is an error, either fix or remove it and document why
If it is real, ask whether your analysis goal can tolerate extreme values
Consider using robust statistics (median, MAD) instead of removing
Report sensitivity: how much do results change with/without outliers?

Robust alternatives

Instead of removing outliers, use statistics that are less sensitive to them. The median is more robust than the mean. Median Absolute Deviation (MAD) is more robust than standard deviation. Robust regression methods like Huber loss can accept outliers without being distorted by them.

The big idea: outliers are questions, not answers. Investigate each one and decide deliberately, rather than scrubbing them reflexively. Sometimes the anomaly is the discovery.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-data-outliers

What is the core idea behind "Outliers: Keep Them, Remove Them, or Investigate?"?
1. A single weird value can distort your entire analysis. But outliers are also where the most interesting stories live. Knowing when to remove them is an art.
2. dataflow
3. Excel and Google Sheets hide a lot of complexity behind a pretty grid.
4. MCAR
Which term best describes a foundational idea in "Outliers: Keep Them, Remove Them, or Investigate?"?
1. IQR
2. outlier
3. z-score
4. robust statistics
A learner studying Outliers: Keep Them, Remove Them, or Investigate? would need to understand which concept?
1. outlier
2. z-score
3. IQR
4. robust statistics
Which of these is directly relevant to Outliers: Keep Them, Remove Them, or Investigate??
1. outlier
2. IQR
3. robust statistics
4. z-score
Which of the following is a key point about Outliers: Keep Them, Remove Them, or Investigate??
1. IQR rule: flag anything below Q1 - 1.5*IQR or above Q3 + 1.5*IQR
2. Z-score: flag anything with |z| > 3
3. Isolation Forest: ML-based anomaly detection, works in high dimensions
4. Domain knowledge: the best detector, often ignored
Which of these does NOT belong in a discussion of Outliers: Keep Them, Remove Them, or Investigate??
1. Isolation Forest: ML-based anomaly detection, works in high dimensions
2. dataflow
3. Z-score: flag anything with |z| > 3
4. IQR rule: flag anything below Q1 - 1.5*IQR or above Q3 + 1.5*IQR
Which statement is accurate regarding Outliers: Keep Them, Remove Them, or Investigate??
1. Check whether the outlier is likely an error (units, typos)
2. If it is an error, either fix or remove it and document why
3. Plot your data first — outliers often jump out visually
4. If it is real, ask whether your analysis goal can tolerate extreme values
Which of these does NOT belong in a discussion of Outliers: Keep Them, Remove Them, or Investigate??
1. dataflow
2. Plot your data first — outliers often jump out visually
3. If it is an error, either fix or remove it and document why
4. Check whether the outlier is likely an error (units, typos)
What is the key insight about "Never blindly remove outliers" in the context of Outliers: Keep Them, Remove Them, or Investigate??
1. A rule of thumb: always investigate before removing. The most embarrassing analyses in history removed outliers that tur…
2. dataflow
3. Excel and Google Sheets hide a lot of complexity behind a pretty grid.
4. MCAR
Which statement accurately describes an aspect of Outliers: Keep Them, Remove Them, or Investigate??
1. dataflow
2. You are analyzing student test scores. 98 percent of scores are between 60 and 100. One score is 3,400,000. Obviously a data-entry error.
3. Excel and Google Sheets hide a lot of complexity behind a pretty grid.
4. MCAR
What does working with Outliers: Keep Them, Remove Them, or Investigate? typically involve?
1. dataflow
2. Excel and Google Sheets hide a lot of complexity behind a pretty grid.
3. Instead of removing outliers, use statistics that are less sensitive to them. The median is more robust than the mean.
4. MCAR
Which of the following is true about Outliers: Keep Them, Remove Them, or Investigate??
1. dataflow
2. Excel and Google Sheets hide a lot of complexity behind a pretty grid.
3. MCAR
4. The big idea: outliers are questions, not answers. Investigate each one and decide deliberately, rather than scrubbing them reflexively.
Which best describes the scope of "Outliers: Keep Them, Remove Them, or Investigate?"?
1. It focuses on A single weird value can distort your entire analysis. But outliers are also where the most interest
2. It is unrelated to foundations workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Outliers: Keep Them, Remove Them, or Investigate??
1. dataflow
2. Three kinds of outliers
3. Excel and Google Sheets hide a lot of complexity behind a pretty grid.
4. MCAR
Which section heading best belongs in a lesson about Outliers: Keep Them, Remove Them, or Investigate??
1. dataflow
2. Excel and Google Sheets hide a lot of complexity behind a pretty grid.
3. Detection methods
4. MCAR

← Back to interactive lesson

Tendril · Creators · AI Foundations