Statistics: The Normal Distribution

If you measure human heights, standardized test scores, manufacturing errors, or just about any natural quantity, the shape of the data often looks the same — a symmetric bell curve peaking near the average and tapering off on both sides. That shape is called the **normal distribution**, or sometimes the Gaussian distribution. It shows up so often in nature and statistics that entire branches of math depend on it.

A normal distribution is fully described by just two numbers:\n\n- The **mean (μ)** — where the curve is centered.\n- The **standard deviation (σ)** — how spread out the data is.\n\nSmall standard deviation: tight, narrow bell. Big standard deviation: wide, flat bell. Same mean, totally different meaning.

For any normal distribution:\n\n- **About 68%** of data falls within 1 standard deviation of the mean.\n- **About 95%** falls within 2 standard deviations.\n- **About 99.7%** falls within 3 standard deviations.\n\nThis is CCSS HSS.ID.4's core claim. Memorize these numbers — they unlock fast estimation.\n\nExample: SAT scores are approximately normal with mean 1050 and standard deviation 200. So:\n- About 68% of scores are between 850 and 1250.\n- About 95% are between 650 and 1450.\n- A score of 1650 is 3 standard deviations above average — rarer than 1 in 400.

A **z-score** tells you how many standard deviations above or below the mean a value is.\n\nFormula: z = (x − μ) / σ\n\n- z = 0 → right at the mean\n- z = 1 → one standard deviation above\n- z = −2 → two standard deviations below\n\nZ-scores let you compare data from different distributions. A 1500 SAT and a 33 ACT are both about z ≈ 2 — meaning both are in the top ~2.5% of test-takers. That's why colleges can compare them.

Not all data is normal. Common non-normal shapes:\n\n- **Skewed right** (income, home prices, reaction times) — long tail to the right, mean > median.\n- **Skewed left** (age at retirement, test scores with easy material) — long tail to the left.\n- **Bimodal** (two peaks) — usually means you're mixing two different populations.\n- **Uniform** — every value equally likely.\n\nIf you apply normal-distribution rules (like the 68-95-99.7 rule) to non-normal data, you'll get nonsense. The first move in any statistical analysis is: **plot the data and look at the shape**.

The normal distribution powers statistical inference — the leap from sample to population. If you measure 100 random people and get a mean height, the distribution of sample means (if you repeated the experiment many times) is approximately normal — even if the underlying data isn't. This is the **Central Limit Theorem**, and it's why polling works, why medical studies work, and why statistics is arguably the most useful branch of applied math in the 21st century.

Understanding the normal distribution is the gateway to all of modern statistics — and increasingly, to data science, machine learning, and quantitative reasoning in almost every field. The bell curve is one of the most powerful ideas ever developed. Learn to use it, and you have a key that unlocks a huge amount of real-world analysis.