A normal distribution is a symmetric, bell-shaped model for a continuous variable. Its center is the mean and its horizontal spread is controlled by the standard deviation Probabilities are areas under the curve, not heights of the curve.
Measure distance in standard deviations
A z-score expresses a value's position relative to its distribution:
Subtracting the mean centers the value; dividing by the standard deviation changes the unit from original measurement units to standard deviations. A positive z-score is above the mean, a negative one is below, and zero is at the mean.
For test scores with mean 70 and standard deviation 8, a score of 82 has It is 1.5 standard deviations above the mean. This interpretation works even when the distribution is not normal, but normal probability calculations require a normal model.
The empirical rule estimates central areas
For a normal distribution, about 68% of observations lie within one standard deviation of the mean, 95% within two, and 99.7% within three.
Symmetry divides the excluded area equally between the two tails. About 5% lie outside two standard deviations, so about 2.5% lie above and 2.5% below
These percentages are approximations for normal or suitably bell-shaped data. Do not apply them automatically to a strongly skewed distribution.
About 95% of a normal distribution lies between z = -2 and z = 2; the two tails share the remaining area.
About 95% of a normal distribution lies between and ; the two tails share the remaining area.
Describe this figure
About 95% of a normal distribution lies between and ; the two tails share the remaining area.
Example: translate an interval into standard units
Suppose package masses are approximately normal with mean 500 g and standard deviation 10 g. Between 480 g and 520 g means within two standard deviations, so about 95% of packages fall there.
Above 510 g means above one standard deviation. Since about 68% lie between 490 and 510, about 32% lie outside, with 16% in the upper tail. Among 2,000 packages, the predicted count above 510 g is about
The model gives an expected proportion, not an exact count in a particular shipment.
Example: use a cumulative probability
A standard-normal table often reports the area to the left. If then the area above 1.5 is
For an interval, subtract cumulative areas: Sketch or mentally locate the requested region first. A right-tail question should not receive a large left-tail probability by accident.
For continuous distributions, the probability of one exact value is zero, so using strict or inclusive endpoints does not change the probability. Real recorded measurements are often rounded; a recorded value then represents an interval.
Reverse a z-score to find an original value
If a question gives a standardized position instead of a measurement, solve the definition for : With mean 70 and standard deviation 8, a value two standard deviations below the mean is
A percentile is a cumulative area, not a z-score. The 90th percentile has 90% of the distribution below it, corresponding to a standard-normal z-score about 1.282. In this score model, it is approximately Do not use 0.90 as the z-score simply because the percentile is 90. Convert the probability to a standardized position first, then return to the original units.
Try it yourself
Heights are modeled as normal with mean 160 cm and standard deviation 6 cm. What z-score corresponds to 148 cm? About what proportion lies above 172 cm?
Answer: Since 172 is two standard deviations above the mean, about 2.5% lie above it by the empirical rule.
What to remember
Standardize with the correct mean and standard deviation, identify the area requested, and distinguish approximate empirical-rule percentages from more precise normal-table or calculator values.