AP Statistics: Percentiles and Cumulative Frequency

Understanding where a single data point stands within a larger group is a fundamental question in statistics. Whether you're interpreting your SAT score, analyzing engineering tolerance limits, or evaluating public health data, percentiles and cumulative frequency provide the precise language for describing relative position. This framework moves beyond the mean and standard deviation to answer the critical question: "How does this one value compare to all the others?"

Defining Percentiles and Percentile Ranks

A percentile is a value on a scale of zero to one hundred that indicates the percentage of observations in a data set that fall at or below that value. The $p^{th}$ percentile is the value below which $p$ percent of the observations lie. It's crucial to distinguish this from a simple percentage. A score of 90% on a test is a percentage correct; a score at the 90th percentile means you scored higher than 90% of the test-takers.

The percentile rank of a specific data value is the percentage of observations in the data set that are less than or equal to that value. For example, if your height of 68 inches has a percentile rank of 72, it means 72% of people in the sample are 68 inches tall or shorter.

Calculating a percentile from raw data involves a few standardized steps:

1. Order the data from smallest to largest.
2. Calculate the index, $i$, using the formula: $i=(\frac{p}{100})\times n$, where $p$ is the desired percentile and $n$ is the number of observations.
3. Find the value:

• If $i$ is an integer, the $p^{th}$ percentile is the average of the data values in positions $i$ and $i+1$.
• If $i$ is not an integer, round $i$ up to the next whole number. The data value at that position is the $p^{th}$ percentile.

Imagine a data set of 11 engineering stress test results (in MPa): {230, 245, 250, 260, 265, 275, 280, 285, 290, 300, 310}. To find the 70th percentile:

• $n=11$, $p=70$, so $i=(70/100)\ast 11=7.7$.
• Since 7.7 is not an integer, round up to 8.
• The value in the 8th position is 285 MPa. Therefore, 70% of the tested materials failed at or below 285 MPa.

Constructing and Interpreting Cumulative Relative Frequency Graphs

A cumulative relative frequency graph (or ogive) is the visual key to understanding percentiles for an entire distribution at a glance. It plots the cumulative relative frequency (the percentile rank) against the corresponding data values.

To construct one from a quantitative data set:

1. Create a frequency table with bins (intervals).
2. Calculate the relative frequency for each bin (frequency / total count).
3. Calculate the cumulative relative frequency for each bin by adding all relative frequencies for that bin and all preceding bins.
4. Plot points using the upper boundary of each bin on the x-axis and the corresponding cumulative relative frequency on the y-axis (as a decimal or percentage).
5. Connect the points with a smooth curve or line segments, starting at the lower boundary of the first bin with a cumulative frequency of 0.

Reading this graph is straightforward but powerful. To find the percentile rank of a value, locate the value on the x-axis, move vertically to the graph, then horizontally to the y-axis to read the cumulative proportion. Conversely, to find the value corresponding to a given percentile (e.g., the median, or 50th percentile), locate the percentile on the y-axis, move horizontally to the graph, then vertically down to the x-axis to read the value. This graph makes the median, quartiles, and other percentiles immediately visible.

Percentiles in Normal Distributions and the z-Score Link

For data that is approximately normally distributed, percentiles can be calculated with precision using the properties of the Normal model and standardized scores. This is where the connection to z-scores becomes essential.

A z-score ($z=\frac{x-\mu }{\sigma }$) tells you how many standard deviations an observation $x$ is from the mean $\mu$. In a Normal distribution, there is a fixed, known relationship between z-scores and percentiles. For instance:

• A z-score of 0 corresponds to the 50th percentile (the mean).
• A z-score of 1 corresponds to approximately the 84th percentile.
• A z-score of -1 corresponds to approximately the 16th percentile.

To find the percentile for a value $x$ in a normal distribution $N(\mu ,\sigma )$:

1. Calculate its z-score.
2. Use a standard Normal distribution table (z-table) or technology to find the area (cumulative probability) to the left of that z-score.
3. This area is the percentile rank, expressed as a proportion. Multiply by 100 to get a percentile.

For example, if SAT Math scores are distributed $N(520,115)$, and you score 650:

• $z=(650-520)/115\approx 1.13$.
• The area to the left of $z=1.13$ in the standard Normal table is about 0.8708.
• Your score of 650 is at approximately the 87th percentile.

The reverse process finds the value corresponding to a given percentile. For the 90th percentile of SAT Math scores:

• Find the z-score where the area to the left is 0.90 (approximately $z=1.28$).
• Convert back to the original scale: $x=\mu +z\sigma =520+1.28(115)\approx 667$.

Thus, a score of about 667 places you at the 90th percentile.

Common Pitfalls

Confusing Percentile with Percentage: This is the most frequent conceptual error. Remember, a percentile is a location within a ranked data set, while a percentage is a fraction out of 100 for a single entity. Scoring 85% correct on a test is not the same as being at the 85th percentile.

Misinterpreting "At or Below": The definition of a percentile includes values equal to the target value. The 75th percentile is the value such that 75% of the data is less than or equal to it. In grouped data or when using certain software algorithms, this distinction can lead to slight variations in calculation, so always know which definition is being applied.

Incorrectly Rounding the Index: When calculating percentiles from raw data using the index formula $i=(\frac{p}{100})\times n$, students often misapply the rounding rule. The rule is clear: if $i$ is not an integer, always round up to the next integer to find the position. Do not round to the nearest integer.

Assuming Normality Without Checking: The elegant z-score to percentile conversion only holds true if the distribution is approximately Normal. Applying it to strongly skewed or non-symmetric data will yield incorrect and misleading percentile estimates. Always check a graph of the data (histogram, Normal probability plot) before using Normal distribution methods.

Summary

• Percentiles describe the relative standing of an observation by giving the percentage of the data at or below that value. The percentile rank is the specific percentage for a given data value.
• A cumulative relative frequency graph is the primary tool for visually estimating percentiles and percentile ranks for any data set, allowing you to move seamlessly from value to rank and vice versa.
• For normally distributed data, z-scores provide a direct bridge to percentiles. The area under the Normal curve to the left of a z-score equals the percentile rank (as a proportion).
• Always clarify the "at or below" definition, round the index calculation correctly, and verify the assumption of Normality before using z-score conversions.