Skewness and Kurtosis - Probability Distributions
Definition
Skewness is a measure of the asymmetry of a probability distribution about its mean. A distribution is:
Positively skewed
- if its right tail is longer or heavier,
Negatively skewed
- if its left tail is longer or heavier,
Symmetric
- if both sides are balanced.
Kurtosis is a measure of the shape of a probability distribution in terms of the peakedness and tail heaviness relative to a normal distribution. It indicates whether the data have:
- a sharper peak and heavier tails,
- a flatter peak and lighter tails,
- or a shape similar to the normal distribution.
Mathematically, skewness and kurtosis are based on central moments of the distribution. If is a random variable with mean and standard deviation , then:
Skewness
Kurtosis
Often, excess kurtosis is used: because the normal distribution has kurtosis equal to 3, so excess kurtosis for a normal distribution is 0.
Main Content
1. Skewness
Meaning and interpretation
- Skewness measures the degree of asymmetry in a probability distribution.
- If a distribution is symmetric, its left and right sides are mirror images, and skewness is approximately zero.
- Positive skewness means the distribution has a long tail to the right. In such cases, most values lie on the lower side, but a few extreme large values pull the mean to the right.
- Negative skewness means the distribution has a long tail to the left. Most values lie on the higher side, but a few extreme small values pull the mean to the left.
Relation among mean, median, and mode
-
In a positively skewed distribution:
-
In a negatively skewed distribution:
-
In a symmetric distribution:
Example of skewness
- Consider exam scores where most students score between 40 and 70, but a few score near 100. This distribution is likely positively skewed.
- Consider retirement age data where most people retire around 60–65, but some retire very early due to illness. This may create left skewness.
ASCII illustration
Symmetric: Positive skew: Negative skew:
/\ /\ /\
/ \ / \ / \
/ \ / \ / \
/ \ / \ / \
--/--------\-- --/--------\____ ____/--------\--
2. Kurtosis
Meaning and interpretation
- Kurtosis describes the shape of the tails and the central peak of a distribution.
- A distribution with high kurtosis tends to have a sharper peak and heavier tails, meaning more frequent extreme deviations.
- A distribution with low kurtosis tends to be flatter with lighter tails, meaning fewer extreme values.
Types of kurtosis
- Leptokurtic: More peaked than a normal distribution and has heavier tails. Extreme values are more common.
- Mesokurtic: Similar to the normal distribution. Excess kurtosis is approximately zero.
- Platykurtic: Flatter than a normal distribution and has lighter tails. Extreme values are less common.
Interpretation of excess kurtosis
- Excess kurtosis > 0: Leptokurtic
- Excess kurtosis = 0: Mesokurtic
- Excess kurtosis < 0: Platykurtic
Example of kurtosis
- A stock market return distribution may be leptokurtic because extreme gains or losses happen more frequently than expected under a normal distribution.
- A uniform-like set of values may be platykurtic because the values are spread evenly and not concentrated around a sharp peak.
ASCII illustration
Leptokurtic: Mesokurtic: Platykurtic:
/\ /\ ____
/ \ / \ __/ \__
/ \ / \ / \
/ \ / \ / \
3. Probability Distributions and Shape Analysis
Why skewness and kurtosis matter in distributions
- Probability distributions describe how probabilities are assigned to possible values of a random variable.
- Two distributions can have the same mean and variance but very different shapes.
- Skewness and kurtosis help capture these shape differences, making distribution analysis more complete.
Connection with the normal distribution
- The normal distribution is symmetric, so its skewness is 0.
- Its kurtosis is 3, and excess kurtosis is 0.
- Many statistical models assume normality, but real data often deviate from it.
- Skewness and kurtosis help identify whether a distribution is far from normal and whether transformations may be needed.
Role in data analysis
- If a distribution is strongly skewed, analysts may apply transformations such as logarithm, square root, or Box-Cox transformation.
- If kurtosis is high, it suggests the presence of outliers or heavy tails, which may require robust methods.
- In hypothesis testing and regression, non-normality can affect confidence intervals and p-values.
Example comparing two distributions
- Distribution A: symmetric, moderate tails, resembles a normal curve.
- Distribution B: same mean as A but right-skewed with heavy tails.
- Without skewness and kurtosis, both may seem similar based on mean and variance alone, but their practical behavior is very different.
Working / Process
1. Collect the distribution data or define the probability model
- Start with either a sample dataset or a theoretical probability distribution.
- Identify the random variable and determine its mean and standard deviation .
- For sample data, compute sample mean and sample standard deviation first.
2. Compute central moments
- Find deviations from the mean: .
- Raise these deviations to the third power for skewness and fourth power for kurtosis.
- Take the expected value of these powered deviations.
- Standardize them by dividing by or so the measures are dimensionless.
3. Interpret the results
- If skewness is near zero, the distribution is approximately symmetric.
- If skewness is positive, the distribution is right-skewed; if negative, left-skewed.
- If excess kurtosis is positive, the distribution has heavier tails than normal.
- If excess kurtosis is negative, the distribution is flatter than normal.
- Use the interpretation to decide whether the data behave normally or whether transformations or robust methods are needed.
Advantages / Applications
Describes distribution shape more completely
- Mean and variance alone do not fully describe a distribution.
- Skewness and kurtosis reveal asymmetry and tail behavior, giving a deeper understanding of the data.
Useful in real-world decision-making
- In finance, skewness and kurtosis help measure investment risk and the chance of extreme losses.
- In quality control, they help detect unusual production behavior or outliers.
- In economics, they help interpret income, wealth, and consumption distributions.
Supports statistical modeling and inference
- These measures help test whether a dataset is approximately normal.
- They guide analysts in choosing suitable transformations and estimation methods.
- They are valuable in comparing probability distributions and evaluating model fit.
Summary
- Skewness shows whether a distribution is symmetric or tilted to one side.
- Kurtosis shows how heavy-tailed or peaked a distribution is compared with normal.
- Both measures help explain the shape of probability distributions beyond mean and variance.
- Important terms to remember: skewness, kurtosis, positive skew, negative skew, leptokurtic, mesokurtic, platykurtic, excess kurtosis