Skewness & Kurtosis Visualizer
Select a distribution below. Each preset uses a mathematically defined family so skewness values are exact. The normal distribution is shown as a grey reference line.
Illustrative distribution — select a preset above
Interpretation
All three distributions below are symmetric (skewness = 0), so you can see how kurtosis changes tail weight independently of asymmetry.
Normal distribution reference
Interpretation
Select any distributions to overlay and compare their shapes. All curves are standardized to the same mean and SD so differences in shape are isolated.
All distributions standardized: mean = 0, SD = 1
| Distribution | Family | Skewness | Excess Kurtosis | Pearson Kurtosis |
|---|
What Is Skewness?
Skewness measures asymmetry in a probability distribution, specifically through the standardized third central moment. When a distribution has a longer or heavier right tail, its skewness is positive. When the left tail is longer or heavier, skewness is negative. A value of zero means no third-moment asymmetry — but that does not make the distribution normal.
The population skewness formula is γ₁ = E[(X−μ)³] / σ³, where μ is the mean and σ is the standard deviation. Cubing the deviations preserves their sign: right-tail extremes produce a positive sum, left-tail extremes a negative one. Dividing by σ³ makes the measure dimensionless, so it applies equally to test scores, financial returns, or any other measurement scale.
An Important Caveat About Mean–Median Ordering
You may have seen the claim that right-skewed data always has mean > median > mode, or that left-skewed data reverses this order. This pattern holds for many common distributions, but it is not a mathematical definition of skewness. Skewness is defined by the third standardized moment, not by the ordering of summary statistics. When working with real data, check the actual distribution shape rather than relying on mean–median ordering alone.
What Is Kurtosis?
Kurtosis is a standardized fourth-moment measure that reflects how much of a distribution's variance comes from extreme observations in the tails. It is calculated as β₂ = E[(X−μ)⁴] / σ⁴. Because deviations are raised to the fourth power, values far from the mean contribute disproportionately to the result. This is why kurtosis is closely tied to tail behavior and sensitivity to outliers.
Kurtosis is often described as a measure of "peakedness," but this is an oversimplification. For some distribution families, peak height and kurtosis move together, but kurtosis is fundamentally a fourth-moment statistic driven by tails and extreme values — not peak geometry alone.
Pearson Kurtosis vs Excess Kurtosis
There are two ways to report kurtosis, and confusing them is one of the most common errors in practice. Always check which convention a tool or software uses.
| Convention | Normal reference | Leptokurtic | Platykurtic | Used by |
|---|---|---|---|---|
| Pearson kurtosis (β₂) | 3 | > 3 | < 3 | Statistics textbooks, older conventions |
| Excess kurtosis (γ₂) | 0 | > 0 | < 0 | R, Python (SciPy), SPSS by default |
The relationship is simple: excess kurtosis = Pearson kurtosis − 3. This tool shows both. When software reports "kurtosis = 0" for a normal distribution, it is using the excess convention. When it reports "kurtosis = 3," it is using the Pearson convention. Neither is wrong — but mixing them up when comparing results across tools produces incorrect conclusions.
The Three Kurtosis Categories
- Mesokurtic: excess kurtosis = 0, Pearson = 3. The normal distribution is the standard mesokurtic reference. A distribution can be mesokurtic without being normal.
- Leptokurtic: excess kurtosis > 0. Higher fourth-moment weight on extremes compared with normal. The Laplace distribution (excess kurtosis = 3) is a clear example. Often associated with heavier tails and greater sensitivity to extreme observations.
- Platykurtic: excess kurtosis < 0. Lower fourth-moment weight on extremes compared with normal. The uniform distribution (excess kurtosis = −1.2) is a useful reference. These distributions still have tails — just lighter ones relative to the normal reference.
Skewness vs Kurtosis
Skewness and kurtosis each describe one aspect of shape but neither captures the full picture. The kurtosis tab in this tool is deliberately built using three symmetric distributions — uniform, normal, and Laplace — so you can see that kurtosis changes independently of skewness. Two distributions can share skewness = 0 and still look nothing alike once you account for tail weight.
| Aspect | Skewness | Kurtosis |
|---|---|---|
| Based on | Third standardized moment | Fourth standardized moment |
| Main concept | Asymmetry of the distribution | Tail and extreme-value behavior |
| Normal reference | 0 | 0 excess / 3 Pearson |
| Sign meaning | Direction of the longer tail | Above or below normal reference |
| Sensitive to outliers | Yes | Very strongly (fourth power) |
| Minimum possible value | No theoretical lower bound | Excess kurtosis ≥ −2 (population) |
Why Moments Do Not Uniquely Determine a Distribution
Two distributions can share the same mean, variance, skewness, and kurtosis while having visibly different shapes. This is why the visualizer labels its curves "illustrative distributions" rather than implying that entering a skewness value generates the one true distribution with those moments. Skewness and kurtosis are useful descriptive tools, but they summarize, rather than fully specify, the distribution.
How to Use This Visualizer
- Skewness tab: Click any preset to see how the distribution shape changes with asymmetry. Enable "Compare with normal" to see the deviation from a symmetric reference.
- Kurtosis tab: All three presets have zero skewness. This isolates tail weight as the only variable changing. Toggle between excess and Pearson kurtosis to understand the two conventions.
- Compare tab: Select multiple distributions and overlay them on the same standardized axes. Use the reference table to see exact theoretical moment values.
- Analyze Data tab: Paste your own numbers to calculate sample skewness and kurtosis. The estimator and convention used are shown explicitly.
Common Misconceptions
Related Pages
Frequently Asked Questions
A positively skewed (right-skewed) distribution has a longer right tail. The bulk of the data clusters toward the left, with a gradual taper toward larger values. The exponential distribution is a classic example, with a theoretical skewness of 2. Income distributions and waiting times often show this pattern. The skewness tab uses exponential-family distributions to demonstrate this shape.
A negatively skewed (left-skewed) distribution has a longer left tail. This is less common in nature but appears in cases like age at a life event with an upper bound, or test scores when most students perform well. This visualizer shows left-skewed examples using a mathematically reflected exponential distribution, which has a theoretical skewness of −2.
Both are correct, depending on the convention. Pearson kurtosis for the normal distribution is 3. Excess kurtosis (Pearson − 3) for the normal distribution is 0. R and Python's SciPy use excess kurtosis by default. Excel's KURT function also returns excess kurtosis. Some older textbooks use the Pearson convention. This tool displays both values and lets you toggle between them on the kurtosis tab.
Not all combinations are mathematically possible. For a probability distribution with finite moments, Pearson kurtosis must satisfy β₂ ≥ γ₁² + 1 (known as the Pearson inequality). This means excess kurtosis ≥ γ₁² − 2. For example, a distribution with skewness = 2 cannot have excess kurtosis below 2. This visualizer avoids this problem by using named, mathematically valid preset distributions rather than arbitrary slider combinations.
Sample kurtosis is highly sensitive to extreme observations because deviations are raised to the fourth power. A single outlier can substantially change the result. Sample kurtosis is also noticeably unstable in small samples — with fewer than 20–30 observations, estimates can swing widely. The data analysis tab reports the sample size alongside the kurtosis value so you can judge how much to trust the result.
Leptokurtic distributions have higher kurtosis than the normal reference (excess kurtosis > 0). They place more fourth-moment weight on extreme deviations, often appearing as heavier tails. The Laplace distribution (excess kurtosis = 3) is a clear example. Platykurtic distributions have lower kurtosis than normal (excess kurtosis < 0). The uniform distribution (excess kurtosis = −1.2) is platykurtic. Neither term means the distribution is good or bad — they simply describe how tail behavior compares to the normal reference.
No. Skewness and kurtosis are two numbers summarizing specific moments of a distribution. Many different distributions can share the same mean, variance, skewness, and kurtosis. These statistics are useful for characterizing shape patterns and comparing distributions, but they cannot replace a full histogram or density plot for understanding data. This is why the visualizer shows full distribution curves rather than just reporting numbers.