Calculating Kurtosis
What you'll learn: How to compute kurtosis values and understand the difference between excess and raw kurtosis.
Now that you know kurtosis measures the "tailedness" of a distribution (from the previous lesson), let's learn how to calculate it.
The Kurtosis Formula
The formula for sample kurtosis is:
Kurtosis = [n(n+1) / ((n-1)(n-2)(n-3))] × Σ[(xᵢ - x̄)⁴ / s⁴] - [3(n-1)² / ((n-2)(n-3))]
Breaking this down in plain English:
- For each data point, find how far it is from the mean
- Raise that difference to the fourth power (versus squared for variance)
- Standardize by dividing by the standard deviation raised to the fourth power
- Apply a correction factor based on sample size
- Adjust with a final subtraction term
Don't worry about memorizing this—software does the calculation for you!
Two Types of Kurtosis
Raw Kurtosis (Fisher's): The complete formula above, which subtracts 3 at the end. A normal distribution has excess kurtosis of 0.
Excess Kurtosis (Pearson's): Sometimes the calculation stops before the final "-3" adjustment. This gives the normal distribution a raw kurtosis of 3.
Think of it like temperature: Celsius (excess kurtosis) sets water's freezing point at 0°, while Kelvin (raw kurtosis) uses a different reference point. Both measure the same thing, just with different baselines.
Most statistical software reports excess kurtosis by default, so:
- Positive values = heavier tails than normal (leptokurtic)
- Zero = normal distribution
- Negative values = lighter tails than normal (platykurtic)
Key Takeaway: Kurtosis uses fourth-power deviations to measure tail weight; excess kurtosis centers the normal distribution at zero by subtracting 3 from the raw value.