Comparing Percentile Positions Across Groups

Comparing Relative Position, Not Just the Scores

A score has meaning partly because of how it compares with other scores in its group. For example, a score of 82 might be unusually high in one group and close to typical in another. Comparing percentile positions helps describe how a score ranks within its group, even when the groups have different distributions or use different scales.

As in Percentiles and Their Interpretation, a percentile describes position within a reference group, not a percentage score. A score at the 80th percentile is at or above about 80% of scores in that group. A percentile rank gives a more direct calculation from observed data: under the at-or-below convention used earlier in this course, it is the percentage of observations less than or equal to the score.

A z-score offers another way to describe relative position. As in Z-Scores: Standardizing a Value, it measures a score’s signed distance from its group’s mean in standard-deviation units. Percentile rank counts the share of observed scores at or below a value; a z-score measures distance from the mean relative to the standard deviation. These are related ideas, but they are not interchangeable.

Comparing Percentile Ranks

To calculate an observed score’s percentile rank, count how many scores in its reference group are less than or equal to it, divide by the group’s size, and multiply by 100%. This uses the at-or-below convention from Percentiles and Their Interpretation. The result may not be a whole number, so report it with suitable rounding.

A score with a higher percentile rank is at or above a larger percentage of its group. That comparison describes relative position; it does not say that the score is higher in raw units, nor that the person is better at every skill. Make sure the groups and variables are relevant to the question. For instance, comparing students’ positions on the same assessment across two classes may be reasonable, while comparing unrelated measures requires care about what “relative standing” is meant to describe.

Percentile ranks depend on the observations in the group. With a small group, one observation can change the calculated percentage substantially. Ties matter too: under the at-or-below convention, every observation tied with the score is counted. State the convention when it helps explain a result.

Worked Example: Comparing Quiz Positions

Comparing Z-Scores

A z-score standardizes a value by subtracting the appropriate group mean and dividing by that group’s standard deviation. Use the mean and standard deviation for the group containing the score, not the other group’s summaries. A positive z-score is above its group’s mean, a negative z-score is below its group’s mean, and zero is at the mean.

A higher z-score means a greater standardized position relative to the group’s mean. It does not always mean “farther above the mean.” For example, \(-1\) is higher than \(-2\), but both scores are below their means; the score with \(z=-1\) is closer to its mean in standard-deviation units. Check the signs as well as the numerical order when interpreting z-scores.

The z-score comparison is useful when scores come from different groups or scales, provided the reference groups are relevant. It describes how many standard deviations above or below each group’s mean a score lies. It does not establish that two people have the same skill or achievement simply because their z-scores match.

Worked Example: Comparing Scores on Two Assessments

Percentile Rank and Z-Score Do Not Say Exactly the Same Thing

Percentile rank is based on the share of observed scores at or below a value. A z-score is based on the value’s distance from its group mean in standard-deviation units. If distributions have different shapes, two scores with the same z-score need not have the same percentile rank. Likewise, a greater z-score in one group does not guarantee a particular percentile-rank difference compared with another group.

The empirical rule, introduced in The Empirical Rule for Mound-Shaped Data, can help relate standard-deviation distances to approximate percentages when a distribution is approximately bell-shaped. Do not use that rule automatically for a distribution that is strongly skewed or otherwise not mound-shaped. Without an appropriate shape assumption, calculate percentile rank from the data if the observations are available, or describe the z-score as a standardized distance rather than assigning it a percentile.

Percentile ranks can also be coarse when groups are small. If a group has 10 observations, each additional observation counted at or below a score changes the percentile rank by 10 percentage points. That limitation is part of the calculation, not a reason to report more precision than the data support.

Worked Example: Ties Change the Percentile-Rank Count

Common Mistakes and AP Exam Tips

• Mixing up a percentile with a percentage score. A score of 80 points is not automatically at the 80th percentile. Calculate or identify its position in the reference group.
• Using the wrong reference group. For a z-score, use the mean and standard deviation from the score’s own group. For a percentile rank, count observations from that group.
• Forgetting the equal-to part of the convention. Under the at-or-below rule, count scores tied with the value as well as scores below it. State that convention when ties affect the answer.
• Calling a higher z-score “farther above the mean.” A higher z-score means a greater standardized position relative to the mean. If both z-scores are negative, both values are below their means; the higher one is closer to its mean in standard-deviation units.
• Claiming a percentile from a z-score without support. A z-score alone does not provide an exact percentile rank for any distribution. Use an appropriate distribution model or the observed group data before making that conversion.
• Overstating what the comparison means. A higher relative position in one group does not prove that a person is better in every respect, that the person would score higher in another group, or that the groups are otherwise comparable.

For a strong response, show the count or the z-score substitution, identify each reference group, and interpret the result in context. When using percentile ranks, report the percentage at or below the score under the stated convention. When using z-scores, state how many standard deviations the score is above or below its own group’s mean. Use “higher standardized position” when comparing z-scores; specify whether the scores are above or below their means rather than assuming that a higher z-score always means farther above.