Kurtosis: Definition and Interpretation
You'll learn what kurtosis measures and how to interpret it when describing data distributions.
What is Kurtosis?
Kurtosis measures how much probability mass sits in the tails (extreme values) and peak of a distribution compared to a normal distribution. While skewness tells you about asymmetry, kurtosis tells you about "tailedness"—whether your data has more or fewer outliers than expected.
Think of kurtosis like describing different hill shapes: some hills have gentle slopes with few rocks at the bottom (light tails), while others have steep peaks and lots of scattered boulders far from the center (heavy tails).
Three Types of Kurtosis
Mesokurtic (kurtosis ≈ 3 or excess kurtosis ≈ 0): Matches the normal distribution. Moderate tails, normal peak.
Leptokurtic (kurtosis > 3 or excess kurtosis > 0): Heavy tails and a sharp peak. More extreme outliers than normal. Think of financial returns—occasional extreme events.
Platykurtic (kurtosis < 3 or excess kurtosis < 0): Light tails and a flatter peak. Fewer extreme values than normal. Think of uniform distributions—values spread more evenly.
Why It Matters
Kurtosis helps you understand risk and variability beyond standard deviation. Two datasets can have identical means and standard deviations but very different tail behavior. High kurtosis warns you: "Expect extreme values more often than the normal distribution suggests."
For data scientists, this affects outlier detection, risk modeling, and choosing appropriate statistical methods.
Key Takeaway: Kurtosis measures tail heaviness and peakedness—high kurtosis means more extreme values (heavy tails), low kurtosis means fewer extremes (light tails), always relative to the normal distribution.