Data visualization

Data visualization turns structured information into visual forms that support perception, reasoning, and communication. A chart assigns data fields to visual encodings such as position, length, angle, area, color, or shape; a map adds geographic position, while an interactive display may add filtering, zooming, and animation.¹ Visualization serves both exploratory analysis, in which viewers search for unexpected relationships, and explanatory communication, in which a selected finding is presented to an audience.

The medium does not remove the need for statistical judgment. Aggregation, binning, normalization, missing values, and the choice of baseline can change the apparent message. A well-designed display therefore states its units, population, time period, and uncertainty clearly. Statistical graphics are especially useful when comparisons are central, because aligned positions and lengths are generally easier to judge than areas or angles.²

Good visualization makes the intended comparison perceptually easy and the underlying evidence inspectable. Cleveland and McGill’s experiments established that people usually estimate aligned position more accurately than length, angle, area, or color saturation; this finding supports bars, dot plots, and line charts for many quantitative comparisons.² Chart selection should also reflect data structure: lines suit ordered sequences, scatterplots expose relationships, maps show spatial variation, and distributions reveal spread rather than only averages.

Design choices can mislead without technically falsifying a value. Truncated axes, uneven intervals, excessive decoration, inappropriate color scales, and hidden denominators may exaggerate differences. Uncertainty intervals, reference lines, direct labels, and annotations can make an analysis more legible. For color, perceptually ordered and color-vision-friendly palettes are preferable; ColorBrewer provides tested schemes for categorical, sequential, and diverging maps.⁶

Modern visualization combines statistical computation, graphical encoding, and sometimes user interaction. Analysts work with libraries and systems that transform tabular, relational, geospatial, or temporal data into views; interactive operations such as brushing, linking, filtering, and details-on-demand can connect overview and inspection.³ Grammar-based approaches describe marks, scales, coordinates, and transformations declaratively, while specialized systems support scientific, cartographic, or business applications.

Interaction is useful when a single static view cannot show the full dataset, but it can also hide exclusions or make a result difficult to reproduce. Accessible practice supplies text alternatives, keyboard access, sufficient contrast, non-color cues, readable labels, and an understandable focus order. The Web Content Accessibility Guidelines provide a widely used framework for such requirements.⁷ Reproducible code, preserved data definitions, and documented transformations help readers audit a visual claim.

The history of visualization includes public-health and administrative work as well as modern computing. Florence Nightingale used a polar-area diagram to argue that preventable disease was a major cause of military deaths, while John Snow used a map of cholera cases and water pumps to investigate an outbreak in nineteenth-century London.¹ These examples show that visualization can function as an argument about institutions and causes, not simply as a neutral display.

Less familiar forms include small multiples, which repeat a consistent graphic across many categories, and glyph-based displays that encode several variables in one symbol. Scientific visualization may represent fields, molecules, astronomical structures, or fluid motion, where occlusion and projection create special interpretive problems. Visualizations can also fail through overplotting, ecological fallacy, unstable color meaning, or a mismatch between map area and population. Their strongest role is not to replace analysis but to expose questions that more formal methods can test.

Visualization quality is evaluated by asking whether viewers can perform the intended task accurately, quickly, and with an appropriate understanding of uncertainty. User studies can compare alternative encodings, while domain experts examine whether aggregation, scales, and annotations preserve the meaning of the data. The same display may be excellent for detecting an outlier but poor for estimating an exact value.

Interpretation also depends on context. A correlation visible in a scatterplot does not establish causation; a dramatic trend may result from changing definitions, sampling, or exposure to the dataset. Readers should inspect the source, denominator, time span, and transformations before drawing conclusions. This combination of perceptual design, statistical literacy, and domain knowledge distinguishes evidence-bearing visualization from ornament.

Visual encoding

The mapping of a data attribute to a visual property such as position, length, color, or shape.

Small multiples

A set of similarly scaled charts that repeat one design across categories or conditions.

Overplotting

The obstruction of marks when many observations occupy the same or nearby visual positions.

Diverging color scale

A color scheme organized around a meaningful midpoint, often used for deviations above and below a reference value.