Normal Distribution Explained
The normal distribution — also called the Gaussian distribution or bell curve — is the most important probability distribution in statistics. It describes how values naturally cluster around a central mean, with values becoming progressively rarer the further they deviate from the centre. The distribution is perfectly symmetric, and its characteristic bell shape emerges whenever a phenomenon is influenced by many independent random factors — which describes an enormous range of real-world measurements.
- Nature's Default Shape: The normal distribution (bell curve) appears naturally in heights, IQ scores, measurement errors, blood pressure, and thousands of other phenomena.
- The 68-95-99.7 Rule: In any normal distribution, 68% of values fall within ±1 SD, 95% within ±2 SD, and 99.7% within ±3 SD of the mean.
- Defined Entirely by Two Numbers: Any normal distribution is completely described by just its mean (μ) and standard deviation (σ).
Why the Bell Curve Emerges
Imagine measuring the height of 10,000 adult men. A few will be very short (below 160 cm). A few will be very tall (above 195 cm). The vast majority will cluster around the average (approximately 175 cm). Plot these measurements on a histogram, and the bars naturally form a bell shape. This happens because height is influenced by hundreds of genetic and environmental factors that each independently push the value slightly up or down. When many independent random effects combine, the Central Limit Theorem guarantees a bell-shaped result.
Normal Distribution: 68-95-99.7 Rule (Mean=0, SD=1)
Standard Deviation Formula
| Symbol | Meaning & Description |
|---|---|
| s | Sample Standard Deviation (using n − 1 denominator) |
| σ | Population Standard Deviation (using N denominator) |
| xᵢ | Individual data values |
| x̄ / μ | Mean (Average) of the dataset |
| n / N | Total number of data points |
Mean = (4 + 7 + 13 + 2 + 8 + 11 + 5) / 7 = 7.1429
Σ(xᵢ - Mean)² = (4 - 7.14)² + (7 - 7.14)² + (13 - 7.14)² + (2 - 7.14)² + (8 - 7.14)² + (11 - 7.14)² + (5 - 7.14)² = 90.8571
Variance = Squared Deviations Sum / n - 1 = 90.8571 / 6 = 15.1429
Standard Deviation = √(15.1429) = 3.8914
The Normal Distribution Formula
The probability density function (PDF) of a normal distribution:
Where:
- μ = mean (centre of the distribution)
- σ = standard deviation (spread)
- e = Euler's number ≈ 2.71828
- π ≈ 3.14159
The formula looks intimidating but encodes a simple idea: probabilities are highest near the mean (where (x − μ)/σ ≈ 0) and fall off exponentially as you move further away.
The 68-95-99.7 Rule (Empirical Rule)
For any normally distributed dataset:
| Range | Percentage of Values |
|---|---|
| Within ±1 SD of mean | 68.27% |
| Within ±2 SD of mean | 95.45% |
| Within ±3 SD of mean | 99.73% |
| More than ±3 SD (outliers) | 0.27% |
Practical example: Adult male heights have μ = 175 cm and σ = 7 cm.
- 68% of men are between 168 cm and 182 cm (175 ± 7)
- 95% of men are between 161 cm and 189 cm (175 ± 14)
- 99.7% of men are between 154 cm and 196 cm (175 ± 21)
The Darts on a Dartboard Analogy
Imagine a skilled dart player throwing at the bullseye blindfolded. Most darts cluster near the centre — they are off-target, but not by much. A few land further away, and very few land far off target. The pattern of dart locations follows a bell curve: most at the centre, progressively fewer as you move outward, symmetrically in all directions. This is the normal distribution — a natural consequence of many small, independent errors or variations combining to produce a final outcome.
- Quality Control: Manufacturing processes aim for parts within a specification range. By knowing the mean and SD of produced parts, engineers can calculate what percentage will be defective (outside spec) using the normal distribution.
- Finance (Risk Modelling): Daily stock returns are approximately normally distributed (though fat-tailed in reality). Risk models use normal distribution to estimate the probability of losses exceeding a threshold (Value at Risk).
- Standardised Testing: SAT, GRE, and IQ tests are designed to produce normally distributed scores. This allows percentile ranks to be calculated precisely — a score at +1 SD corresponds to the 84th percentile.
- Medical Reference Ranges: Lab test "normal ranges" (e.g., haemoglobin, blood pressure, blood glucose) are typically defined as μ ± 2σ — capturing 95% of the healthy population.
- Assuming Everything Is Normal: Many real-world distributions are not normal — income is right-skewed, insurance claims are heavily right-tailed, earthquakes follow a power law. Applying normal distribution formulas to non-normal data produces incorrect probabilities.
- Confusing the Standard Normal with All Normals: The standard normal distribution has μ = 0 and σ = 1. Any normal distribution can be transformed to standard normal using z-scores, but they are not the same thing.
- The "Fat Tails" Oversight: Real financial returns have "fat tails" — extreme events (market crashes, Black Swan events) occur far more frequently than a normal distribution would predict. This caused catastrophic failures in financial models during the 2008 crisis.
- Using ±2 SD as a Hard Boundary: The 95% range is probabilistic, not absolute. If a medical test's normal range is μ ± 2σ, then 5% of perfectly healthy people will fall outside it — generating false positives.
ANormal Distribution (Symmetric)
A normal distribution is perfectly symmetric around its mean. Mean = Median = Mode. The tails on both sides are identical mirror images. Normal distributions arise from sums of many independent random variables (Central Limit Theorem). They are the foundation of classical statistics: t-tests, ANOVA, regression, and confidence intervals all assume approximate normality.
BSkewed Distribution (Asymmetric)
Skewed distributions have one tail longer than the other. Right-skewed (positively skewed) distributions (like income or stock returns) have mean > median. Left-skewed distributions (like age at retirement) have mean < median. Skewness occurs when an asymmetric force limits values on one side (e.g., income cannot be negative, but can be arbitrarily high). Skewed data requires different statistical methods — log transformations, non-parametric tests, or median-based summaries.
When Can You Assume Normal Distribution?
| Phenomenon | Approximately Normal? | Notes |
|---|---|---|
| Heights of adults (same sex) | ✅ Yes | Classic example |
| IQ scores | ✅ Yes | By design (standardised to N(100, 15)) |
| Exam score distributions | ✅ Often | Depends on test design |
| Daily temperature anomalies | ✅ Approximately | Seasonal patterns must be removed first |
| Household incomes | ❌ No | Heavily right-skewed; use log-normal |
| Stock market daily returns | ⚠️ Approximately | Fat tails exist in reality |
| Wait times / service times | ❌ No | Typically exponential or gamma distribution |
| Manufacturing defect counts | ❌ No | Poisson distribution for rare events |
Z-Score Calculator Sandbox
Tweak variables below to see the formula calculate instantly.
Z-Score Calculator
Calculate the z-score (standard score) of a raw value, given the population mean and standard deviation. Includes probability percentile lookup.
Related Educational Guides
What Is a Percentage? Core Math Calculations Explained
Understand what percentages are, how to solve the three core percentage problem types, and why percentages are the universal language of proportion.
What Is Standard Deviation? Variance & Data Spread
Understand standard deviation, variance, and the 68-95-99.7 rule — the primary statistical measure of data spread used in finance, science, and quality control.
Mean vs. Median vs. Mode: Comparing Central Tendencies
Learn the three measures of central tendency, when to use each, and why using mean for skewed data leads to misleading conclusions.