Z-score Calculator

A positive score lies above the mean, a negative score lies below it and zero equals the mean. Magnitude describes standardized distance, not practical importance.

The calculation does not require normality. Normal-distribution percentages or tail probabilities do require an appropriate distribution assumption, which this page does not test or calculate.

Choose the standard deviation that belongs with the reference data. Mixing an observation from one population with a mean or spread from another produces a number with little meaning.

Outliers can change the mean and standard deviation, and robust summaries may be better for strongly skewed data. Inspect the data rather than using a z-score as the only quality check.

A standard deviation of zero means all reference values are identical and standardization is undefined. That is why the endpoint requires a positive value.

Record enough context to reproduce this single-value z-score standardization: input source, unit, date and any method choices. A bare number is easy to misread when it is copied into a report or compared later.

Round for communication only after the calculation. The endpoint formatting improves readability but does not increase the accuracy of the measurements, assumptions or source data.

When the mean and standard deviation are estimated from the same sample, some statistical procedures use studentized residuals or t distributions rather than treating a simple z-score as a known-population statistic. The correct method depends on the inference being made.

A percentile conversion under normality uses the standard normal cumulative distribution. This page deliberately stops at z because adding a percentile would imply a distributional assumption that may not fit the data.

Standardizing reverses cleanly: x = μ + zσ. Use that identity to check an answer. For z = 1.5, mean 70 and standard deviation 10, the reconstructed value is 70 + 1.5×10 = 85.

Comparing z-scores from two datasets is meaningful only when each reference distribution is suitable for its observation. Different testing populations, years or instruments can make apparently comparable standardized values answer different questions.

For quality-control measurements, a large absolute z-score can prompt investigation, but control limits and process capability require their own assumptions and methods. One isolated standardized value does not establish that a process is out of control.

If the reference statistics are rounded, the z-score inherits that rounding. Published means and standard deviations with only one decimal place cannot support a highly precise four-decimal interpretation, even though the endpoint formats four decimals.

For repeated observations from one person, population reference z-scores and within-person change scores are different. A result standardized against population variability does not directly show whether that person changed by a meaningful amount.