The Normal Distribution and The Empirical Rule

In our last lesson, we looked at the probability of single, isolated events—flipping a coin, clicking an ad, or a server going offline.

But what happens when you zoom out? What happens when you plot thousands, or even millions, of independent data points on a single chart?

If you measure the height of every woman in a city, or the weight of every apple in an orchard, or the exact dimensions of 10,000 factory-made screws, a strange and beautiful thing happens. The chaotic, random data stops looking random. It organises itself into a perfect, symmetrical curve.

This is called the Normal Distribution, but you probably know it as the Bell Curve.

It is the most important shape in statistics. Once you realise your data follows this shape, predicting the future becomes incredibly easy.

Anatomy of the Bell Curve

The Normal Distribution isn’t just a curve; it is a mathematical law of nature. For a dataset to be considered perfectly “Normal,” it must follow these strict rules:

1. Perfect Symmetry: The left side of the curve is an exact mirror image of the right side. (If you remember the lesson on Skewness, a true Normal Distribution has a skew of exactly $0$).
2. The Holy Trinity: The Mean, the Median, and the Mode are all the exact same number. They all sit dead center at the absolute peak of the bell.
3. Asymptotic Tails: The tails of the curve stretch out forever in both directions. They get closer and closer to the bottom axis, but mathematically, they never actually touch it.

Here is the brilliant part: Every single normal distribution in the universe is defined by just two numbers.

• The Mean ($\mu$): This tells the curve where to sit. It anchors the center of the bell.
• The Standard Deviation ($\sigma$): This tells the curve how fat or skinny to be. A small standard deviation makes a tall, squeezed spike. A large standard deviation makes a short, wide, flattened hill.

The Magic Trick: The 68-95-99.7 Rule

Why do statisticians obsess over the Normal Distribution? Because of a mathematical cheat code known as the Empirical Rule.

If you know that your data forms a bell curve, you instantly know exactly where every piece of data lives without having to measure them all individually. You just count outward from the center using standard deviations as your measuring stick.

Here is the rule:

• 68% of all your data will fall within 1 Standard Deviation of the mean. (This is the fat middle of the bell).
• 95% of all your data will fall within 2 Standard Deviations of the mean.
• 99.7% of all your data will fall within 3 Standard Deviations of the mean.

Anything beyond 3 standard deviations is incredibly rare. In data science, this is where we draw the mathematical line for an “extreme outlier.”

The Rule in the Real World

Let’s look at adult male heights in the United States. They form a perfect normal distribution.

• The Mean is roughly 5 feet 10 inches (70 inches).
• The Standard Deviation is 3 inches.

Using the Empirical Rule, we can instantly predict the population without looking at a single person:

• 68% of men are between 5’7″ and 6’1″ (1 standard deviation up and down).
• 95% of men are between 5’4″ and 6’4″ (2 standard deviations).
• 99.7% of men are between 5’1″ and 6’7″ (3 standard deviations).

If you meet a man who is 6’8″, you don’t need a complex formula to know he is an extreme outlier. He is more than 3 standard deviations away from the mean, meaning he falls into that tiny, $0.3\%$ tail of the curve.

How to Calculate Normal Distributions in Software

While the Empirical Rule gives you a perfect shortcut for exactly 1, 2, or 3 standard deviations, what if you need to know the probability of a man being exactly 71.5 inches tall? To find the exact area under the curve, you can use these simple software functions.

(Assume the Mean is 70 and the Standard Deviation is 3).

Microsoft Excel

Excel uses the NORM.DIST function to find the probability of a value falling below a specific point (the cumulative area to the left).

• Find the probability of a value being less than 71.5: Type =NORM.DIST(71.5, 70, 3, TRUE)
• Note: The “TRUE” command tells Excel to measure the entire area up to that point, which is almost always what you want.

Python

In Python, we use the norm function from the scipy.stats library. The .cdf() (Cumulative Density Function) measures the area to the left of your value.

from scipy.stats import norm

mean_height = 70
std_dev = 3
target_height = 71.5

# Calculate probability of being shorter than 71.5 inches
probability = norm.cdf(target_height, loc=mean_height, scale=std_dev)

print(f"Probability: {probability:.4f}") # Output: 0.6915 (69.15%)

R

R uses the pnorm() function natively to calculate the exact same cumulative area under the curve.

mean_height <- 70
std_dev <- 3
target_height <- 71.5

# Calculate probability of being shorter than 71.5 inches
probability <- pnorm(target_height, mean = mean_height, sd = std_dev)

print(probability) # Output: 0.6914625

See It in Action

Use this interactive tool to visualise exactly how changing the Mean and Standard Deviation morphs the curve, and watch how the Empirical Rule perfectly locks the data into those 68-95-99.7 percentages.

The Empirical Rule is incredible, but it only works if you are comparing things on the same scale. What happens if you want to compare a 6’8″ man to a 40kg dog? In the next lesson, we will learn how to standardise the entire universe using Z-Scores.