What Is Descriptive Statistics?

Descriptive statistics is the branch of statistical analysis concerned with summarizing and presenting the features of a data set without generalizing beyond the observations in hand. It produces numerical summaries such as means, standard deviations, and quantiles, along with tables and graphics that make the shape of the data visible. The defining boundary is one of scope: a descriptive summary makes a claim about the sample that was measured, whereas an inferential procedure makes a probabilistic claim about a wider population from which the sample was drawn.

The methods predate modern inference by centuries, growing out of demographic and actuarial record keeping, and they were substantially extended in the 1970s by the exploratory data analysis movement associated with John Tukey. In engineering practice they serve as the first pass over any measured data set, since the choice of a downstream model depends on what the summary reveals about center, spread, shape, and anomalies.

Measures of Center and Spread

The arithmetic mean, median, and mode each answer the question of a typical value differently, and the difference matters when a distribution is skewed or contains outliers. The mean is efficient for symmetric distributions but is pulled by extreme observations, while the median has a breakdown point of 50 percent and is preferred for income, latency, and lifetime data. Spread is described by the variance and standard deviation, by the range, and by robust alternatives such as the interquartile range and the median absolute deviation. The NIST/SEMATECH Engineering Statistics Handbook, a reference written specifically for scientists and engineers, treats these summaries as the quantitative half of a two-part examination of any measurement process, the other half being graphical.

Shape, Position, and Graphical Summaries

Beyond center and spread, descriptive work characterizes shape through skewness and kurtosis, and position through percentiles, quartiles, and z-scores. Graphical summaries carry much of the load in practice. Histograms and kernel density estimates show modality, run sequence plots expose drift and shifts over time, lag plots reveal autocorrelation, and normal probability plots test the distributional assumption that later methods depend on. The box plot, introduced by Tukey, encodes the median, the quartiles, and whiskers extending to the most extreme points within 1.5 interquartile ranges of the hinges, with points beyond that flagged as candidate outliers. Chapter 1 of the NIST/SEMATECH e-Handbook, on exploratory data analysis, organizes these techniques around four underlying assumptions that measured data are often expected to satisfy: random drawing, fixed distribution, fixed location, and fixed variation.

Multivariate and Categorical Summaries

Descriptive work extends past a single variable. Contingency tables and cross-tabulations summarize categorical data through counts, proportions, and marginal totals. Pairwise relationships are described by covariance, the Pearson correlation coefficient, and rank-based alternatives such as Spearman's rho, all of which describe association within the sample without asserting causation. Scatterplot matrices and parallel coordinate plots extend graphical description to several dimensions at once. A discussion of how these summaries are applied to real observational data sets appears in the NCBI Bookshelf chapter on exploratory data analysis in electronic health records, which stresses screening for missingness and implausible values before any modeling begins.

Applications

Descriptive statistics has applications in a range of fields, including:

• Statistical process control and manufacturing yield monitoring
• Reliability engineering and failure-time summary
• Signal and image analysis for noise characterization
• Clinical trial and epidemiological baseline reporting
• Network and system performance measurement
• Survey research and official statistics reporting
• Machine learning feature engineering and data quality screening