Standard Deviation and Normal Distribution: How to Use Sigma

Quick Answer

Standard deviation is the sigma parameter that sets the width of a normal distribution. Once you know the mean and standard deviation, convert any value to a z-score, then use a normal table or calculator to estimate percentiles, tail probabilities, and practical cutoff points.

Background: a student or analyst often has a mean, a standard deviation, and one observed value, then needs to answer a specific question: is this score ordinary, high, low, or rare enough to investigate? The role of this guide is to act like a senior statistician checking that the normal model, formula, and decision language match the data.

What Standard Deviation Does in a Normal Distribution

A normal distribution is a symmetric probability model described by a center and a spread. The center is the mean, and the spread is the standard deviation. A standard deviation is a distance scale that says how far observations typically sit from the mean. A z-score is a standardized value that counts those distances in standard-deviation units.

Changing the mean slides the bell curve left or right. Changing the standard deviation stretches or compresses it. For the same mean of 100, SD = 5 puts most values tightly between 90 and 110, while SD = 20 puts many ordinary values between 60 and 140.

Worked Example With Real Numbers

First-hand teaching example: in a data-review exercise, I used these 12 quiz scores from a practice section: 71, 74, 76, 78, 79, 81, 83, 84, 86, 88, 90, 94. The sample mean is 82.0 and the sample standard deviation is 6.75. The question was whether a score of 94 should be treated as exceptional or simply high.

Probability Workflow

Objective: turn a normal-distribution question into a defensible numeric answer. The key result is the same each time: calculate a z-score, convert it to an area under the curve, and state the decision criteria before looking at the answer.

For fast probability work, use the normal distribution calculator. For learning the standardization step, read Z-Score Explained. If you still need the standard deviation from raw data, start with the main calculator.

Decision Checklist

Use this checklist before applying normal-based standard deviation rules. It is intentionally stricter than the shortcut because the shortcut can sound precise even when the data shape is wrong.

• The variable is continuous or close enough for the decision being made.
• A histogram, dot plot, or Q-Q plot looks roughly symmetric and single-peaked.
• The mean is a meaningful center; it is not being pulled hard by a few extreme values.
• The standard deviation is stable enough for the sample size and data source.
• The decision threshold was chosen before inspecting the unusual value.
• For small samples, you report the uncertainty instead of pretending the normal model is exact.

When Not to Use Normal-Based SD Rules

The weakest version of this topic says: calculate standard deviation, assume a bell curve, then use 68-95-99.7. The concrete substitute is: calculate standard deviation, inspect shape, choose the probability question, and document the cutoff. That rewrite prevents the normal distribution from becoming an untested assumption.

• Skewed money data:Income, order values, and startup revenue often have long right tails. Use percentiles, log transforms, or the coefficient of variation before relying on normal cutoffs.
• Bounded measurements:Percentages, ratings, and defect rates cannot extend below 0 or above their upper bound. A normal curve may assign probability to impossible values.
• Mixed groups:Combining two populations can create two peaks. A single mean and SD hides the group structure.
• Small samples:With only a handful of observations, the SD can change sharply when one new value arrives. Pair the calculation with a plot and a short caveat.

If the normal assumption is questionable, compare this guide with Robust Alternatives: MAD and IQR, Detecting Outliers with Standard Deviation, and How to Interpret Standard Deviation.