The Normal Distribution: The 68-95-99.7 Rule

Heights, measurement errors, sample means — again and again, data piles up in a symmetric bell around a center. That shape is the normal distribution, and it sits underneath more of statistics than any other curve.

Note

The normal distribution is defined by just two numbers: the mean (where it centers) and the standard deviation (how wide it spreads).

The 68-95-99.7 Rule

In any normal curve, the proportion of values within a given number of standard deviations of the mean is fixed — as the shaded bands above show.

About 68% of values fall within 1 standard deviation of the mean
About 95% fall within 2 standard deviations
About 99.7% fall within 3 — a value beyond that is genuinely rare

This is why a z-score of 3 raises eyebrows: under normality, only 0.3% of values live that far out.

Why It Matters

Z-scores, confidence intervals, t-tests, and much of regression lean on normality — of the data, or of the sample means. The Central Limit Theorem is the reason it shows up so often: averages of many independent values tend toward normal even when the raw data isn't.

Tip

You often don't need the raw data to be normal — just the sampling distribution of the mean, which the Central Limit Theorem hands you for free at large sample sizes.

Checking Normality

Histogram: is it symmetric and bell-shaped, or skewed?
Q-Q plot: do the points hug the diagonal line?
Shapiro-Wilk test: a formal check, though it's touchy on large samples

Watch out

Don't assume normality — skewed or heavy-tailed data breaks methods that rely on it. Check first, then choose the test.

Normality Checks in SKARI

When you run a test in the Statistical Lab, SKARI checks normality for you and flags when it fails — steering you to a Welch correction or a nonparametric alternative instead of quietly returning an invalid result.

Automatic normality checks before parametric tests
Distribution previews in the Data Editor's column profiling
Suggested alternatives when the assumption doesn't hold

Takeaway

The assumption is checked for you — so you don't build a conclusion on a curve your data never followed.

Frequently Asked Questions

What's a z-score?

How many standard deviations a value sits from the mean. It turns any normal distribution into a common scale.

Does my data have to be normal?

Often only the sample mean does, thanks to the Central Limit Theorem — but heavy skew or small samples still call for care.

What if it isn't normal?

Transform the data, use Welch's correction, or switch to a nonparametric test that doesn't assume normality.

Conclusion

The normal distribution is the backbone of statistical inference. Know the 68-95-99.7 rule, check the assumption, and you'll know when to trust it — and when to reach for something else.

Takeaway

Two numbers describe the whole curve — and the 68-95-99.7 rule tells you where nearly everything lands.

Descriptive Statistics

Center, spread, and shape

Data Normalization

Z-scores in practice

Hypothesis Testing Basics

Where normality is used