Percentiles and z-scores answer a common statistical question: how a single observation stands relative to others. They help compare performances within one distribution and across distributions measured on different scales.

Relative position in a distribution

In statistics, a relative position tells where one value falls compared with the rest of a group. A raw score alone is often not enough. A score of 82 could be excellent in one class, average in another, and weak in a third, depending on the overall distribution.

When statisticians compare relative position, they ask questions such as:

• How much of the data is below this observation?

• How far is this observation from the center?

• Is this value unusually high or low for its own data set?

One way to describe position is with a percentile.

Percentiles depend on rank rather than on the actual numerical gaps between observations. That makes them useful even when the distribution is skewed or when the scale of measurement differs across groups.

A value at the higher percentile has a higher standing in its own distribution. If one student is at the 88th percentile and another is at the 73rd percentile, the first student scored at or above a larger share of the comparison group. The key idea is not the raw score itself, but the position of that score among all observations.

Standardizing with z-scores

Another way to compare relative position is with a z-score, which uses the mean and standard deviation of the distribution.

Unlike percentiles, z-scores describe distance from the mean in standardized units.

A normal (bell-shaped) curve marked at $-3\sigma, -2\sigma, -1\sigma, m, 1\sigma, 2\sigma, 3\sigma$. The vertical reference lines emphasize that z-scores measure standardized distance from the mean in units of the standard deviation. Source

This can be especially useful when two data sets use different numerical scales.

For a quantitative value, the z-score is found by standardizing the distance from the mean.

When sample statistics are used, $\mu$ may be replaced by $\bar{x}$ and $\sigma$ by $s$.

The sign and size of a z-score carry the meaning:

• Positive z-score: the value is above the mean.

• Negative z-score: the value is below the mean.

• z-score near 0: the value is close to the mean.

• Larger absolute value of z: the value is farther from the mean and more unusual relative to that distribution.

Comparing positions within one data set

Within a single distribution, both percentiles and z-scores can rank observations.

Percentiles are often easiest when the main goal is to describe standing. A higher percentile means a stronger position in the ordered data. This is especially natural for test scores, growth charts, or any context where rank matters more than exact distance.

Z-scores are more informative when the distance from the mean matters. Two observations may have nearby percentiles but noticeably different z-scores if one lies much farther from the mean in standard deviation units. In that case, the z-score gives more detail about how unusual the value is.

When comparing two points in the same data set:

• The observation with the higher percentile has more observations below it.

• The observation with the higher z-score is relatively farther above the mean.

• If both measures point to the same value as higher, the comparison is straightforward.

Comparing positions between different data sets

Comparisons across different data sets are common in AP Statistics. Raw values alone can be misleading because the centers, spreads, and scales may differ.

Percentiles allow comparison by rank within each group. If one observation is at the 91st percentile in one data set and another is at the 84th percentile in a different data set, the first observation has the stronger relative standing in its own group.

Z-scores allow comparison by standardized distance from the mean. If one value has $z=1.4$ and another has $z=0.8$, the first value is farther above its group mean relative to that group’s spread. This makes z-scores especially helpful when data sets are measured in different units or have different standard deviations.

However, the two measures are not identical in meaning:

A normal distribution annotated with standard-deviation tick marks and multiple aligned scales, including z-scores and percentile equivalents. It helps students see how rank (percentile) and standardized distance (z-score) correspond on a bell-shaped model, while still representing different kinds of information. Source

• Percentiles compare rank.

• z-scores compare distance from the mean after standardizing.

Because of this, the better measure depends on the question being asked.

Choosing between percentiles and z-scores

Use percentiles when:

• the goal is to compare rank or standing,

• the context naturally describes performance by percent below,

• or the distribution is irregular enough that ranking is more meaningful than distance from the mean.

Use z-scores when:

• the goal is to compare how far values are from their means,

• the data are quantitative,

• and the distributions are similar enough that mean and standard deviation give useful summaries.

A percentile does not tell how tightly packed the data are near that value. A z-score does not tell what proportion of observations are below the value unless extra distribution information is available. For that reason, percentiles and z-scores answer related but different questions about relative position.

Interpreting comparisons in context

On AP Statistics problems, interpretation matters as much as computation. A correct comparison should always name the measure being used and connect it to the data set.

Strong contextual statements usually:

• identify whether the comparison is based on percentile or z-score,

• state which observation has the higher relative position,

• and explain what that means within the relevant group.

Avoid comparing raw scores from different settings without standardizing or using rank. Also avoid saying that one value is "better" unless the context supports that word. In statistics, the precise claim is usually that one value is higher relative to its distribution.