Historical Context & Motivation
The need to summarize and interpret numerical data is as old as organized society itself, but the formal mathematical machinery we now call descriptive statistics coalesced over several centuries of intellectual effort. Early census records from Babylon, Egypt, and China quantified populations and harvests, yet these inventories lacked any notion of a single representative value for a collection of measurements. The conceptual leap from raw tabulation to a concise numerical summary—what we now term a statistic—required both the development of probability theory and the practical demands of astronomy, where reconciling slightly discrepant observations of the same celestial event was a matter of scientific credibility.
The central question these pioneers addressed remains the same question you face on the DAT: given a set of numerical observations, how do we best characterize its center and its spread? The concepts of mean, median, variance, and standard deviation provide the definitive answers, and understanding their properties is essential for interpreting experimental data, evaluating research findings, and solving quantitative reasoning problems under timed conditions.
Core Principles & Definitions
Descriptive statistics can be organized around two complementary ideas: measures of central tendency, which identify a typical or representative value in a dataset, and measures of dispersion, which quantify how far individual observations deviate from that center. Together, these two families of statistics provide a compact yet informative summary of any univariate distribution. A solid grasp of the following five foundational concepts will cover the vast majority of DAT quantitative reasoning items related to data analysis.
Mean (Arithmetic Average)
Median
Mode
Variance
Standard Deviation
Visual Explanation — Central Tendency vs. Spread
The diagram below illustrates two datasets that share the same mean (μ = 50) but differ markedly in their standard deviations. Dataset A (σ = 5) concentrates its observations near the center, producing a tall, narrow bell curve, while Dataset B (σ = 15) distributes observations across a wider range, yielding a flatter, broader curve. This visual comparison underscores a critical principle: the mean alone is an incomplete summary. Two distributions can have identical centers yet convey entirely different information about the variability inherent in their underlying processes.
Notice that approximately 68% of observations in any normal distribution fall within one standard deviation of the mean. For Dataset A, this range is 45–55, whereas for Dataset B it extends from 35–65. This 68-95-99.7 rule (also called the empirical rule) states that roughly 68%, 95%, and 99.7% of data lie within 1, 2, and 3 standard deviations of the mean, respectively. While the DAT may not always present perfectly normal data, this rule provides a powerful mental benchmark for assessing whether a given observation is typical or extreme.
Mathematical Framework
With the conceptual groundwork established, we now formalize each statistic algebraically. Pay close attention to the distinction between population parameters (denoted by Greek letters) and sample statistics (denoted by Latin letters), as the DAT occasionally tests this notational convention. The sample formulas use n − 1 in the denominator (Bessel's correction) to produce an unbiased estimate of the population variance.
Detailed Breakdown — Effect of Outliers and Skew
One of the most frequently tested concepts on the DAT is how outliers and skewness differentially affect the mean and median. Because the mean incorporates every value through summation, a single extreme observation can drag it toward the tail of a distribution, whereas the median, being a positional measure, remains anchored at the center of the ordered dataset. This robustness property makes the median the preferred measure of central tendency for income distributions, home prices, and many biological measurements that exhibit right skew.
The relationship between mean and median also serves as a quick diagnostic for skewness. In a right-skewed (positively skewed) distribution the mean exceeds the median because the right tail drags the balance point upward. Conversely, in a left-skewed (negatively skewed) distribution the mean is less than the median. When the two measures coincide, the distribution is symmetric. This heuristic—mean vs. median as a skewness indicator—appears regularly on the DAT and should become an automatic check whenever you encounter a dataset.
| Skew Direction | Mean vs. Median | Tail Direction |
|---|---|---|
| Right (positive) skew | Mean > Median | Long tail extends to the right |
| Symmetric | Mean ≈ Median | Tails are approximately equal |
| Left (negative) skew | Mean < Median | Long tail extends to the left |
Worked Example — Full Descriptive Analysis
A researcher measures the diastolic blood pressure (mmHg) of 8 patients in a clinical trial: 72, 78, 80, 82, 85, 88, 90, 95. Compute the mean, median, variance, and standard deviation of this sample.
Strengths & Limitations of Each Measure
No single statistic is universally superior; each has contexts in which it excels and situations in which it misleads. The DAT expects you to select the most appropriate measure for a given scenario, so understanding the trade-offs below is as important as knowing the formulas.
| Measure | Strengths | Limitations |
|---|---|---|
| Mean | Uses every data point; algebraically tractable; minimizes the sum of squared deviations; essential for computing variance and standard deviation. | Sensitive to outliers; can misrepresent highly skewed distributions; not defined for ordinal or nominal data. |
| Median | Resistant to outliers; appropriate for skewed data; applicable to ordinal data; splits data into equal halves. | Ignores the magnitude of most values; less amenable to algebraic manipulation; less efficient estimator when data are symmetric. |
| Mode | Applicable to categorical data; identifies peaks; can indicate multimodality in a distribution. | May not exist, or may not be unique; carries little information about spread; unstable in small samples. |
| Variance | Additive for independent variables; foundational for ANOVA, regression, and many inferential procedures. | Expressed in squared units, making direct interpretation unintuitive; sensitive to outliers (even more so than the mean). |
| Standard Deviation | In the same units as the data; directly interpretable via the empirical rule; widely reported in scientific literature. | Still sensitive to outliers; not additive (variances are, but SDs are not); assumes interval or ratio scale. |
Connection to Inferential Statistics & Advanced Theory
Descriptive statistics serve as the gateway to inferential statistics, where sample statistics are used to draw conclusions about population parameters. The Central Limit Theorem guarantees that the distribution of sample means approximates a normal distribution as n grows large, with mean μ and standard deviation σ/√n (the standard error). This result transforms the sample mean from a mere summary into a test statistic that can be compared against hypothesized values, forming the basis of confidence intervals and hypothesis tests.
| Descriptive Concept | Inferential Extension |
|---|---|
| Sample mean (x̄) | Point estimator of the population mean μ; used in z-tests, t-tests, and ANOVA. |
| Sample variance (s²) | Unbiased estimator of σ²; F-tests compare two variances; chi-square tests evaluate a single variance. |
| Standard deviation (s) | Used to compute the standard error SE = s/√n, the building block of confidence intervals. |
| Median | Foundation for nonparametric tests (Wilcoxon, Mann-Whitney) when normality cannot be assumed. |
While the DAT focuses primarily on descriptive computations, a conceptual understanding of how these quantities extend into inference strengthens your problem-solving intuition. For example, knowing that the standard error decreases with √n explains why larger samples yield more precise estimates—a principle that occasionally surfaces in DAT questions about experimental design and data interpretation. Similarly, recognizing that the variance has the mathematical property of additivity for independent random variables (Var(X + Y) = Var(X) + Var(Y)) helps explain why researchers prefer variance over standard deviation in theoretical derivations, even though standard deviation is more interpretable in applied reporting.
Practice Problems
Summary
Descriptive statistics organize raw data into concise numerical summaries. The mean (arithmetic average) captures the balance point of a distribution and serves as the foundation for the variance and standard deviation, which quantify how tightly observations cluster around that center. The median provides a robust alternative when outliers or skewness distort the mean, and comparing the two reveals the distribution's asymmetry. For samples, always apply Bessel's correction (n − 1) when computing variance and standard deviation.
On the DAT, these concepts translate into concrete skills: computing each statistic by hand, selecting the appropriate measure for a given data scenario, interpreting the empirical rule (68-95-99.7) for normally distributed data, and reasoning about how changes to a dataset (adding outliers, increasing sample size) affect the mean, median, and standard deviation. Mastering these fundamentals also establishes the conceptual vocabulary required for inferential statistics—confidence intervals, hypothesis tests, and regression—where sample statistics serve as estimators of unknown population parameters.