In statistics, the "moments" of a distribution describe its shape. While they might sound complicated, every moment is fundamentally just an average (expected value) of the data raised to a certain power. Adjust the data points below to see how this works in practice.
In mathematics and statistics, a moment is a quantitative measure of the shape of a set of points. The generalized \( n \)-th moment of a variable \( X \) evaluated around a reference center \( c \) and scaled by a factor \( a \) is the expected value (average) of \( \left(\frac{X - c}{a}\right)^n \). By changing \( n \), \( c \), and \( a \), we can derive the mean, variance, skewness, and kurtosis!
* Deviations from the grey Normal Reference curve indicate Skewness (leaning left/right) and Kurtosis (abnormally sharp peaks or fat tails).
This table reveals the "under the hood" calculations. It shows how each raw data point is mathematically transformed before being summed up. Notice how the final Average (Σ/N) row perfectly matches the moment values below!
| Point | Raw Value \( x_i \) |
Deviation \( x_i - \mu \) |
Dev² \( (x_i - \mu)^2 \) |
Std. Dev ( \( z_i \) ) \( \frac{x_i - \mu}{\sigma} \) |
Std. Dev³ \( z_i^3 \) |
Std. Dev⁴ \( z_i^4 \) |
|---|
The mean is the average of the raw values. It tells us the "center of mass" of the distribution.
Variance is the average of the squared differences from the mean. It tells us the "scale" or how spread out the data is. Unlike higher moments, we do not divide by standard deviation here. Why? Because variance is exactly what we use to measure that scale! If we standardized the data first, the variance would always just equal 1.
Skewness is the average of the cubed differences. We divide by standard deviation (σ) to standardize it, making it a "scale-free" measure of asymmetry. Because it's cubed, signs are preserved: negative outliers pull it left, positive pull it right.
Kurtosis is the average of the data raised to the fourth power. Like skewness, we divide by standard deviation (σ) to remove units (like meters or dollars) so we can strictly measure the "tailedness" or shape of the distribution. Extreme outliers get massively amplified.