Percentiles & Quartiles — Percentiles and Quartiles

The need to describe where a particular observation sits relative to others in a dataset is nearly as old as statistics itself. When Francis Galton studied human variation in the late nineteenth century, he realized that measures of central tendency alone—means and medians—were insufficient for characterizing the full picture of how individual measurements compared to one another. What was needed was a systematic way to express relative standing: the proportion of a distribution that falls below a given value. This motivation gave rise to the concepts we now call percentiles and quartiles, two of the most widely used tools in descriptive statistics.

At their core, percentiles and quartiles answer a deceptively simple question: where does a particular value sit in the overall distribution? Whether you are interpreting a patient's blood-pressure reading against population norms, evaluating a student's GRE score, or identifying outliers in a financial dataset, the machinery of percentiles and quartiles provides the formal framework. The remainder of this lesson develops that framework rigorously, from definitions and computation to interpretation and application.

Before computing any percentile or quartile, one must first arrange the data in ascending order. Every concept in this lesson operates on ranked (sorted) data. With that prerequisite in mind, the following foundational ideas define the landscape of positional statistics.

A well-crafted diagram can make the abstract notion of 'splitting a distribution' immediately concrete. The following SVG illustrates a continuous bell-shaped distribution partitioned at the three quartile boundaries. Each shaded region contains approximately 25% of the total area under the curve, corresponding to the four quarters of the data.

Several features of this diagram deserve attention. First, notice that the quartile boundaries are not evenly spaced along the horizontal axis—Q₁ and Q₃ are closer to Q₂ than they are to the tails, reflecting the fact that data in a bell-shaped distribution are densest near the center. Second, while this diagram uses a symmetric distribution for clarity, quartiles are equally well defined for skewed distributions; in a right-skewed distribution, Q₃ would be pushed further from the median than Q₁, and the IQR would shift accordingly. Finally, the shaded regions are areas under the density curve, so each represents a probability (or proportion) of 0.25, reinforcing that quartiles are conceptually tied to cumulative distribution functions.

Computing percentiles from raw data requires a precise algorithm. Multiple methods exist (there are at least nine recognized variants), but the locator method presented here is widely taught in introductory statistics courses and provides a clear, step-by-step procedure. We also present the percentile rank formula for completeness.

Quartiles form the skeleton of the five-number summary, which consists of the minimum, Q₁, median (Q₂), Q₃, and maximum. This summary provides a compact yet informative description of a dataset's center, spread, and range. When plotted graphically as a box-and-whisker plot (or simply box plot), the five-number summary becomes a powerful visual diagnostic for skewness, outliers, and distributional shape.

In a box plot, the length of each whisker relative to the box immediately communicates skewness. If the right whisker (from Q₃ to max) is substantially longer than the left whisker (from min to Q₁), the distribution is right-skewed. If the median line sits closer to Q₁ than to Q₃ inside the box, this further confirms a rightward stretch. Outlier conventions typically flag any observation beyond 1.5 × IQR from the nearest quartile as a potential outlier. In the diagram above, 1.5 × 11 = 16.5, so the lower fence would be 7 − 16.5 = −9.5 and the upper fence would be 18 + 16.5 = 34.5; since all data fall within these fences, no outliers are present.

Consider the following dataset of exam scores (already sorted in ascending order) for a class of 15 students: 42, 48, 53, 56, 60, 63, 67, 70, 73, 78, 82, 85, 88, 92, 97. We will compute the 40th percentile (P₄₀), the three quartiles (Q₁, Q₂, Q₃), and the percentile rank of the score 82.

Percentiles and quartiles are enormously useful, but like every statistical tool they carry assumptions and trade-offs. Understanding when to rely on positional measures versus moment-based measures (like the mean and standard deviation) is essential for choosing the right descriptive summary for a given analysis.

The concepts developed in this lesson are not merely descriptive conveniences; they connect directly to more advanced topics that you will encounter in subsequent coursework. Percentiles are intimately related to the cumulative distribution function (CDF), and quartiles serve as the foundation for several inferential and diagnostic techniques.

Perhaps the most elegant theoretical link is between percentiles and the CDF. For a continuous random variable X with CDF F(x), the p-th quantile is defined as Q(p) = F⁻¹(p), the value x such that F(x) = p. The sample percentile you compute from data is therefore an estimator of this population quantity. As sample size n → ∞, the sample percentiles converge to the true population percentiles—a manifestation of the Glivenko–Cantelli theorem, which states that the empirical CDF converges uniformly to the true CDF. This asymptotic guarantee is what makes percentile-based descriptive statistics theoretically sound, not just practically convenient.