DataTerminology.comNEW
Moment-based statistic describing the tail heaviness and outlier-proneness of a distribution relative to normal.
DESCRIPTION
Kurtosis quantifies tail weight in probability distributions using the fourth standardized moment. Positive excess kurtosis indicates heavy tails (leptokurtic), while negative indicates light tails (platykurtic). It complements variance and skewness in shape analysis. In practice, kurtosis helps detect outliers and assess model assumptions. Interpretation varies by definition (e.g., excess vs. Pearson’s kurtosis).
ALTERNATIVE DEFINITIONS
Wikipedia
In probability theory and statistics, kurtosis is a measure of the ‘tailedness’ of the probability distribution of a real-valued random variable.
Centers on tail behavior as the defining characteristic.
NIST
Kurtosis measures the heaviness of the tails of a distribution relative to a normal distribution and is based on the fourth central moment.
Highlights the moment-based definition and comparison to normal.
RELATED TERMS
Moments Normal Distribution Heavy-Tailed Distribution Leptokurtic Platykurtic Mesokurtic Variance Outliers
COMPETING TERMS
KNOWLEDGE AREAS
Mathematics Data Science Data Quality
TARGET AUDIENCE
Data Scientist Data Analyst BI Specialist
EXTERNAL SOURCES
- Wikipedia - Kurtosis
- Penn State Eberly College of Science - Lesson on distribution shape and kurtosis
- Wolfram MathWorld - Mathematical definition and properties
TOP QUESTIONS
What does high Kurtosis indicate?
High kurtosis suggests heavy tails and a higher likelihood of extreme outliers compared to a normal distribution.
Category: Concept
What is excess Kurtosis?
Excess kurtosis subtracts 3 from Pearson’s kurtosis so the normal distribution has zero excess kurtosis.
Category: Definition
When is Kurtosis useful in practice?
When assessing model assumptions, risk of outliers, or suitability of mean/variance-based methods.
Category: Relevancy
What are common misunderstandings about Kurtosis?
It measures tail weight, not peakedness alone; interpretations vary with definition used.
Category: Misunderstanding
How can Kurtosis mislead analysis?
Small samples yield unstable estimates; extreme outliers can dominate the metric.
Category: Risk
How do you reduce Kurtosis in data?
Use robust methods, transformations, trimming, or winsorization to reduce tail impact.
Category: How-to
What are typical Kurtosis values for common distributions?
Normal has excess 0; Laplace is leptokurtic (>0); Uniform is platykurtic (<0).
Category: Example
How does Kurtosis relate to skewness?
Both are shape descriptors: skewness measures asymmetry, kurtosis measures tail weight.
Category: Concept
What estimators exist for Kurtosis?
Use moment-based sample estimators with bias corrections for finite samples.
Category: Requirements