Binary, Nominal, Ordinal, and Continuous Variables: How Should You Describe Your Data?

Is a Variable Binary, Nominal, Ordinal, or Continuous?

Researchers routinely describe variables using labels such as binary, nominal, ordinal, and continuous. The labels seem simple until a dataset contains values such as 0 and 1, a five-point satisfaction item, age in years, education levels coded 1 through 6, or a count of publications.

Which of these are categorical? Which are quantitative? Does assigning numbers to categories change their type? And where do discrete variables fit?

The central principle is straightforward: classify a variable according to the information represented by its possible values. Numerical coding alone tells you remarkably little. A spreadsheet is quite capable of making three different measurement structures look suspiciously alike.

Four Common Variable Types at a Glance

Start With What the Values Mean

Binary variables have two categories

A binary or dichotomous variable has two possible categories. Examples include yes/no, present/absent, completed/did not complete, treatment/control, and positive/negative.

The CDC describes dichotomous variables as nominal variables with two mutually exclusive categories. Penn State likewise treats binary variables as a form of categorical data.

Binary variables are frequently coded 0 and 1, but those numbers generally function as category codes. The coding can be computationally convenient, yet it does not mean that the substantive variable has become continuous.

Nominal variables have categories without an intrinsic order

A nominal variable classifies observations into categories that do not have an inherent quantitative ranking.

Examples might include blood type, academic program, country of residence, mode of transportation, or research methodology category.

If academic programs are coded engineering = 1, business = 2, and education = 3, the numbers remain labels. Education is not quantitatively “greater than” business because its code is larger.

NIST describes a nominal scale as classification without intrinsic ordering among the classes. That property is more important than whether the categories are represented by words, letters, or numbers.

Ordinal variables add meaningful order

An ordinal variable also consists of categories, but those categories have a meaningful ranking.

Examples include disease stage, educational attainment, class standing, agreement categories, or severity classifications such as mild, moderate, and severe.

What ordinal measurement does not guarantee is equal spacing. The CDC notes that ordinal values can be ranked but are not necessarily evenly spaced. NIST similarly emphasizes that ordinal scales support ordering, while differences between values are not necessarily comparable.

This is why coding low = 1, medium = 2, and high = 3 does not establish that the difference between low and medium is quantitatively identical to the difference between medium and high.

Continuous variables represent quantities along a continuum

A continuous variable is quantitative and can, conceptually, take any value within a range. Examples include height, weight, duration, blood pressure, and many physical measurements.

The CDC defines continuous variables as having the potential for an infinite number of values along a continuum. In practice, instruments record finite precision, but that does not necessarily alter the conceptual type of the underlying variable.

The broader continuous-versus-categorical distinction is useful because these kinds of variables generally require different descriptive and analytical treatment.

Binary is usually a special case, not a separate level of measurement

Binary, nominal, ordinal, and continuous are frequently listed together for practical convenience, but they do not occupy exactly the same conceptual level.

A binary variable tells you that there are two possible categories. Nominal and ordinal tell you something about the relationship among categories. A binary variable can therefore be nominal, and in some settings an ordering of its two states may also be substantively meaningful.

Penn State notes that for binary variables such as pass/fail, the distinction between nominal and ordinal treatment may make little practical difference for many purposes.

It is therefore more precise to think of binary as describing the number of categories rather than as a wholly separate measurement scale parallel to nominal and ordinal.

Discrete quantitative variables complete an important missing piece

Where does “number of children” fit? It is not nominal, ordinal, or continuous in the usual sense.

It is a discrete quantitative variable. Its values have numerical meaning, but the possible values are countable. Number of publications, hospital visits, errors, courses completed, or citations within a defined period are similar examples.

Penn State distinguishes discrete quantitative variables from continuous ones: counts take countable values, whereas continuous quantities such as elapsed time can vary along a continuum.

Nominal and ordinal describe more than whether categories use numbers

Consider educational attainment coded as:

• 1 = less than high school;
• 2 = high school;
• 3 = two-year degree;
• 4 = bachelor's degree;
• 5 = master's degree;
• 6 = doctoral or professional degree.

The values are ordered, but the numerical difference between adjacent codes does not necessarily represent an equal amount of educational attainment. Penn State therefore classifies this kind of variable as categorical rather than quantitative.

Compare that with number of courses completed. Values 1, 2, 3, and 4 represent actual counts, and the difference between one and two courses has the same numerical interpretation as the difference between three and four. The second variable is quantitative and discrete.

Identical-looking integers can therefore encode very different information.

What about Likert-type responses?

An individual Likert-type item, such as strongly disagree, disagree, neither agree nor disagree, agree, and strongly agree, consists of ordered categories. At the item level, it is therefore naturally described as ordinal.

Questions become more complicated when several items are combined into a scale score. Researchers sometimes analyze such composite scores using methods ordinarily applied to quantitative data, but that practice involves assumptions about the measurement properties of the resulting score. It should not be justified merely by saying that the responses were coded 1 through 5.

If several measures are combined, the resulting variable may also raise separate questions about what constitutes a composite variable and how the combination is justified.

A variable's representation can change

Suppose age is originally recorded in years and months. Researchers might later create categories such as 18–24, 25–34, and 35–44.

The first representation is quantitative and may be treated as continuous for many purposes. The second is an ordinal categorical variable because the age groups have a meaningful order.

The underlying characteristic is still age, but the representation has changed. This is why researchers should distinguish the construct or characteristic being studied from the variable actually entered into the analysis.

Creating categories from a continuous measure may sometimes serve a substantive or practical purpose, but it also changes the information available. The implications are considered more directly when deciding whether a continuous variable should be categorized.

Variable type affects what summaries are meaningful

Nominal variables are commonly summarized using frequencies and proportions. Ordinal variables can additionally support ordered summaries such as cumulative percentages. Quantitative variables can support numerical summaries of magnitude, although the appropriate measures depend on their distributions and measurement properties.

The CDC, for example, distinguishes frequency distributions for nominal and ordinal variables from summaries of continuous variables using measures of central location and spread.

Variable type also affects graphical display and model selection. A bar chart of categories answers a different question from a histogram of continuous measurements, just as logistic regression for a binary outcome serves a different modeling purpose from ordinary linear regression for an appropriately modeled continuous outcome.

Classifying Variables in a Student Survey

Now imagine the researcher codes academic program as 1, 2, 3, and 4 and proficiency as 1 through 4. Both columns contain integers, but neither has become a continuous measurement. Academic program remains nominal; proficiency remains ordinal.

Meanwhile, number of courses also contains integers, yet those integers are genuine counts. The crucial difference is what the numbers represent.

Common Mistakes When Describing Data Types

How Should You Describe a Variable in Your Study?

Work from the possible values and their meaning. A variable codebook should allow another researcher to understand not only what numbers or labels appear in the dataset but what those values represent.

Then document any transformations separately. If age was collected continuously but analyzed in categories, say so. If a five-item instrument was summed into a composite score, explain the scoring procedure rather than describing the original items and resulting score as though they were necessarily the same type of variable.

Before Describing Your Data Types

Questions About Binary, Nominal, Ordinal, and Continuous Data

Describe What the Values Mean, Not Just How They Look

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.