A z score describes how far a particular value is from the mean of a distribution, measured in standard deviations. A z score of 2 means the value is two standard deviations above the mean; a score of −1 means it is one standard deviation below it.
What a z score tells you
A z score, also called a standard score, puts a measurement into context by comparing it with the distribution’s mean and standard deviation. It tells you both the direction and the distance of the difference:
- A positive z score means the value is above the mean.
- A negative z score means the value is below the mean.
- A z score of zero means the value equals the mean.
- The score’s absolute size indicates how many standard deviations separate the value from the mean.
For example, suppose test scores have a mean of 70 and a standard deviation of 10. A score of 80 has a z score of 1, while a score of 60 has a z score of −1. Both are one standard deviation from the mean, but on opposite sides. The z score does not, by itself, say whether either score is good, bad, unusual, or important; that depends on the setting and the distribution.
How to calculate a z score
For a value in a population with a known mean and standard deviation, use:
z = (x − μ) / σ
Here, x is the value being assessed, μ is the population mean, and σ is the population standard deviation. Subtracting the mean centers the value around zero; dividing by the standard deviation expresses the difference in standard-deviation units.
Using the test-score example, a score of 80 gives (80 − 70) / 10 = 1. If you have only a sample and use its sample mean and sample standard deviation, the calculation is often written z = (x − x̄) / s. The choice of formula depends on what data you have and the purpose of the analysis. Some inferential procedures use a t statistic rather than a z statistic when a population standard deviation is unknown, especially with small samples.
Interpreting z scores and percentiles
A z score can help compare values measured on different scales. If one student’s height and another student’s test score each have a z score of 1.5 in their respective comparison groups, each value is 1.5 standard deviations above its group mean. That does not make the raw measurements equivalent, but it does show their relative positions within those distributions.
When data follow a normal, bell-shaped distribution, z scores can also be related to percentiles and probabilities. In a standard normal distribution, a z score of about 1.96 is near the 97.5th percentile, and about 95% of values lie between −1.96 and 1.96. These are model-based interpretations, not universal facts about every dataset. A z score alone does not establish a percentile unless the distribution or a suitable reference distribution is known.
A useful distinction is that a z score measures distance in standard deviations, while a percentile describes the proportion of observations at or below a value. They answer related but different questions.
When z scores are useful
Z scores are used in statistics, education, research, and quality control when people need a common way to express relative position. Common uses include:
- Comparing measurements: Standardized values can make results from scales with different units easier to compare, provided each value is judged against an appropriate distribution.
- Finding unusual observations: A large absolute z score can flag a value for review. It is a clue, not proof of an error or a genuine outlier.
- Using normal-distribution tables: A z score locates a value on the standard normal curve, helping estimate probabilities under the normal model.
- Standardizing data: Analysts may transform measurements to have a mean of zero and a standard deviation of one, for tasks such as preparing variables for certain statistical or machine-learning methods.
For a direct comparison to be meaningful, the reference groups and measurement conditions should also make sense. A score standardized against one age group, school, or time period may not be comparable to a score standardized against another.
Limitations and common misunderstandings
The formula works whenever the standard deviation is greater than zero, but interpretation requires care. If every observation has the same value, the standard deviation is zero and the z score is undefined because division by zero is impossible.
Another common misconception is that a z score of 2 always means a value is rare. It is relatively far from the mean in standard-deviation units, but rarity depends on the distribution’s shape. In a normal distribution, a z score of 2 is around the 97.7th percentile. In a skewed or heavy-tailed distribution, the proportion of values beyond that point may differ substantially.
Likewise, the often-used guideline that values beyond ±2 or ±3 are potential outliers is only a screening convention, not a universal rule. Check the data, the measurement process, and the distribution before deciding what an extreme score means. Outliers can be mistakes, valid unusual cases, or evidence that the model does not fit the data well.
Frequently Asked Questions
Can a z score be negative?
Yes. A negative z score means the value is below the mean of its reference distribution. For example, −1.2 means it is 1.2 standard deviations below the mean.
What does a z score of 1 mean?
It means the value is one standard deviation above the mean. It does not automatically mean the value is in the 84th percentile; that interpretation is approximately true only when the reference distribution is normal.
Is a z score the same as a percentile?
No. A z score gives distance from the mean in standard-deviation units. A percentile gives the proportion of observations at or below a value. A distribution is needed to convert between them.
What is a good z score?
There is no universally good z score. Its meaning depends on what is measured and the comparison group. A high score may be desirable on one measure and undesirable on another.
Can z scores be used when data are not normally distributed?
Yes, they can still describe how far a value is from the mean in standard-deviation units. However, normal-curve probability and percentile interpretations may not apply, particularly for skewed or heavy-tailed data.
The key idea
A z score is a standardized description of a value’s position relative to a mean: it states how many standard deviations above or below that mean the value lies. To interpret it well, identify the reference group, understand the distribution, and avoid treating a large or positive score as meaningful without considering what the measurement represents.
