Z-Scores: Standardizing Data for Comparison

A z-score tells you how many standard deviations a given value sits from the mean of its distribution. A z-score of 0 sits exactly at the mean. A z-score of 1 sits one standard deviation above the mean. A z-score of -1.5 sits one and a half standard deviations below the mean. Once you've expressed a value as a z-score, you can compare it directly to values from any other distribution, look up its probability in a standard normal table, or use it as a standardized input to other statistical calculations.

This guide covers what a z-score is and how to calculate one, what z-scores mean and how to interpret them, common z-score thresholds and their probabilities, the relationship between z-scores and the standard normal distribution, how researchers use z-scores in practice, the limitations of the method, and how to report z-scores in APA format. For the broader framework that z-scores sit within, see Editor World's complete guide to descriptive statistics. For the reference distribution z-scores are anchored to, see the guide to the normal distribution and its properties.

What Is a Z-Score?

A z-score is a standardized version of a raw value. Instead of expressing the value in its original units (dollars, points, milliseconds), the z-score expresses it as a distance from the mean, measured in standard deviations. This puts every value on a common scale, which makes comparison and probability calculation straightforward.

Z-scores are sometimes called standard scores or normal scores. The term "z" comes from the convention of using the letter z to denote a variable that follows the standard normal distribution. That distribution is the special case of the normal where the mean is 0 and the standard deviation is 1. Z-scores express raw values in those standard-normal terms, which is what makes them useful.

The Z-Score Formula

The formula for a z-score is simple. Subtract the mean from the value, then divide by the standard deviation.

z = (x − μ) / σ

where x is the raw value, μ is the population mean, and σ is the population standard deviation. When working with a sample rather than a population, use the sample mean and sample standard deviation in their place, often written as M and SD in research contexts. The two formulas produce the same kind of z-score; the only difference is whether the mean and standard deviation are population parameters or sample estimates.

The standard deviation underlies the formula in a meaningful way. The z-score expresses how a value compares to the typical spread of the distribution. If the standard deviation is small, even modest distances from the mean produce large z-scores. If the standard deviation is large, the same distance produces a small z-score. For a deeper look at the standard deviation and what it represents, see Editor World's guide to standard deviation and variance.

A Worked Example

IQ scores are designed to follow a normal distribution with a mean of 100 and a standard deviation of 15. Suppose a person scores 130 on an IQ test. What's their z-score?

z = (130 − 100) / 15 = 30 / 15 = 2.0

The z-score is 2.0, which means the score sits exactly two standard deviations above the mean. About 97.5% of the population has a lower IQ. The z-score of 2.0 carries the same information as the raw score of 130, but it makes the position of the value within the distribution immediately interpretable. Anyone familiar with the standard normal distribution can read a z-score of 2.0 and know it represents an unusually high value.

Now consider a person who scores 92 on the same test. Their z-score is:

z = (92 − 100) / 15 = -8 / 15 ≈ -0.53

A z-score of -0.53 means the score sits a little more than half a standard deviation below the mean. About 30% of the population has a lower IQ. The negative sign signals that the value is below the mean, and the absolute value tells you how far. With practice, researchers read z-scores fluently without needing a calculator or a probability table.

Comparing Values from Different Distributions

The most useful application of z-scores is comparing values from different distributions. Raw scores from two different measures aren't directly comparable. A score of 50 on one test could be excellent or terrible, depending on the test. Z-scores translate raw scores onto a common metric, which makes comparison straightforward.

Consider a study using two different anxiety measures. The Beck Anxiety Inventory (BAI) has a sample mean of 12 with a standard deviation of 8. The State-Trait Anxiety Inventory (STAI) has a sample mean of 38 with a standard deviation of 10. Participant 1 scores 20 on the BAI and 50 on the STAI. At raw values, the two scores look unrelated. After standardization, the comparison is direct:

BAI z = (20 − 12) / 8 = 1.0
STAI z = (50 − 38) / 10 = 1.2

Both z-scores sit just above 1, meaning the participant's anxiety levels on the two scales are comparable: each is about one standard deviation above the sample mean. The raw scores looked very different. The standardized scores reveal that the two scales are giving consistent signals about this participant's anxiety. This kind of standardization is routine when researchers want to compare across measures that use different units.

Common Z-Score Thresholds and Probabilities

Certain z-score values correspond to specific probabilities in the standard normal distribution. Memorizing the most common thresholds saves time when interpreting z-scores in real work.

• z = ±1. About 68% of values fall within this range. This is the one-standard-deviation interval and corresponds to the 68 in the 68-95-99.7 rule.
• z = ±1.645. About 90% of values fall within this range. Used for one-tailed tests at the 0.05 significance level.
• z = ±1.96. About 95% of values fall within this range. Used for two-tailed tests at the 0.05 significance level and for 95% confidence intervals.
• z = ±2. About 95.4% of values fall within this range. Close enough to the 1.96 threshold that many researchers use ±2 as a quick approximation.
• z = ±2.58. About 99% of values fall within this range. Used for two-tailed tests at the 0.01 significance level and for 99% confidence intervals.
• z = ±3. About 99.7% of values fall within this range. Values beyond an absolute z-score of 3 are uncommon and often warrant attention as potential outliers.

These thresholds are derived from the standard normal distribution, which gives them a precise mathematical meaning. When you see a z-score of 1.96 in a research paper, it almost always refers to a 95% confidence interval or a two-tailed significance threshold.

How Z-Scores Are Used in Research

Z-scores show up in several common research workflows.

Identifying outliers

Z-scores provide a numerical criterion for identifying outliers. A common rule of thumb flags any value with an absolute z-score above 3 as a potential outlier. Some researchers use 2.5 or 3.29 as the threshold, depending on field conventions and how conservative the analysis needs to be. The rule isn't a hard decision; flagged values should be inspected to determine whether they're errors, valid but unusual cases, or normal variation that happens to fall in the tail.

Standardizing variables for regression

When predictors in a regression are measured in different units, the regression coefficients aren't directly comparable. A coefficient of 0.5 on income (in thousands of dollars) and a coefficient of 0.5 on age (in years) describe very different magnitudes of effect. Standardizing the predictors (converting each one to z-scores) puts all predictors on a common scale, which lets you compare the standardized coefficients to assess relative importance. This is the source of "standardized beta coefficients" in regression output.

Constructing composite scores

When researchers combine multiple measures into a composite (an overall stress score derived from several questionnaires, for example), the measures often use different scales. Standardizing each measure to z-scores before summing or averaging puts them on equal footing. Otherwise, the measure with the largest raw-score range dominates the composite.

Reporting position within a normative distribution

Many standardized assessments report results as z-scores or as transformations of z-scores (T-scores, IQ scores, SAT scores). These transformations all anchor a raw value to a known distribution, which lets clinicians and researchers communicate position without specifying the underlying scale every time.

Foundation for inferential statistics

Z-scores are the backbone of many inferential tests. The z-test for proportions, the test statistic in large-sample mean comparisons, the formula for confidence intervals, and the conversion of test statistics to p-values all depend on z-scores. Anyone learning inferential statistics learns z-scores first because every other formula builds on them.

Z-Scores When Data Isn't Normal

A common misunderstanding is that z-scores require the underlying data to be normally distributed. They don't. Z-scores can be calculated for any data with a mean and standard deviation, regardless of distribution shape. The calculation itself doesn't depend on normality.

What does depend on normality is the probability interpretation of z-scores. The 68-95-99.7 rule and the standard normal probability tables only apply when the underlying distribution is approximately normal. If your data is heavily skewed or has unusual tail behavior, a z-score of 2 doesn't necessarily correspond to the 97.5th percentile. The z-score still expresses position in standard deviation units, but the percentile interpretation requires checking the actual distribution shape.

For deeper coverage of how to detect non-normality and what it means for your analysis, see Editor World's guide to skewness and kurtosis.

Reporting Z-Scores in APA Format

APA convention italicizes z when used as a statistical symbol in text, formulas, or in reporting test statistics. Z-scores themselves are typically reported to two decimal places.

For an individual z-score: "The participant scored 130 on the IQ test (z = 2.00), placing them in the top 2.5% of the population."

For a z-test result: "The proportion was significantly different from the hypothesized value, z = 2.84, p = .005."

For standardized regression coefficients, the term "standardized beta" or the symbol β with explicit labeling is used rather than calling it a z-score. The underlying mechanism is the same: each predictor has been converted to z-score form before the regression is run.

Common Mistakes With Z-Scores

• Forgetting the sign. A z-score of -2 and a z-score of 2 are equally far from the mean in absolute terms, but they're on opposite sides. When interpreting results, always check whether the value is above or below the mean.
• Assuming the percentile interpretation always holds. The 68-95-99.7 rule applies to normal distributions. For skewed or heavy-tailed data, the same z-score corresponds to a different percentile.
• Confusing z-scores with t-scores. Z-scores use the population (or known) standard deviation. T-scores use a sample-based standard deviation and follow the t-distribution rather than the standard normal. For small samples, the t-distribution has heavier tails than the standard normal, which affects probability calculations.
• Standardizing the wrong way for composite scores. When combining multiple measures into a composite, each measure should be standardized using its own mean and standard deviation. Standardizing all measures using a single pooled mean and standard deviation distorts the composite.
• Treating large z-scores as automatically meaningful. A z-score of 3 in a sample of 30 might just be the largest value in a small dataset. A z-score of 3 in a sample of 10,000 is much more unusual. Sample size affects how much weight to give large z-scores.
• Using z-scores on data that should be analyzed differently. Ordinal data, count data, and bounded proportions often need specialized methods rather than z-score-based analysis. Standardizing without thinking about whether the analysis is appropriate produces numbers that look meaningful but aren't.

When Professional Editing Helps

Z-score reporting is one of the smaller places in a statistics manuscript where notation errors slip through. Missing italics on z, wrong decimal precision, confused references between z-scores and t-scores, and percentile claims that assume normality without checking are all common in first drafts. Editor World's academic editing services include review of statistical notation, APA compliance, and the substantive accuracy of methodological claims. The same standard is applied across dissertation editing, journal article editing, and essay editing. 100% human editing, no AI at any stage. You choose your own editor from verified profiles, and a free sample edit is available before you commit. Browse available editors by subject expertise to find someone whose background matches your field.

Frequently Asked Questions About Z-Scores

Content reviewed by Editor World editorial staff. Editor World, founded in 2010 by Patti Fisher, PhD, graduate of The Ohio State University, provides professional editing and proofreading services for academic researchers, doctoral candidates, faculty, business professionals, and authors worldwide. BBB A+ accredited since 2010 with 5.0/5 Google Reviews and 5.0/5 Facebook Reviews. More than 100 million words edited for over 8,000 clients in 65+ countries. Stevie Award winner: Gold 2019, Bronze 2018 and 2025. Native English editors from the United States, the United Kingdom, and Canada with subject-matter expertise across the social sciences, the natural and physical sciences, medicine, engineering, computer science, and the humanities. 100% human editing, no AI at any stage. Less than 5% of applicants are accepted to the editor panel. Recommended by the Boston University Economics Department, University of San Diego, University of Michigan, UCLA, University of Missouri, and more.