Types of Data and Variables

Before you can analyze data, you need to know what kind of data you have. The type of data determines which graphs, summary statistics, and analytical methods are appropriate. Applying the wrong technique to the wrong data type produces results that look numerical but mean nothing.

This page covers the three most important classification systems: categorical vs quantitative, discrete vs continuous, and the four levels of measurement.

Categorical vs Quantitative Data

Every variable you encounter falls into one of two broad categories.

Categorical (qualitative) data describes a quality, label, or grouping. The values are names or categories, not numbers you perform arithmetic on. Even when categorical data uses numbers (like zip codes or jersey numbers), those numbers are labels — averaging them would be meaningless.

Examples: blood type, eye color, job title, zip code, marital status, brand preference.

Quantitative (numerical) data represents a measurable quantity. You can add, subtract, and average these values, and the results are meaningful.

Examples: height, temperature, number of patients, test score, annual income, weight.

The key test: Can you meaningfully calculate a mean? If yes, the data is quantitative. If the “average” would be nonsensical (like an average blood type), it is categorical.

Example 1: Classifying Hospital Data

A hospital patient record includes these fields. Classify each as categorical or quantitative.

Notice that room number is the classic trap. Just because a value contains digits does not make it quantitative. Ask: does arithmetic on this number produce something meaningful? Adding room 302 and room 115 gives 417, which is not a useful result.

Discrete vs Continuous Variables

Quantitative data divides further into two subtypes based on what values are possible.

Discrete variables have countable values, usually whole numbers. You can list every possible outcome. There is no value “between” two consecutive counts — you cannot have 2.7 children or 14.3 defective items on a production line.

Examples: number of children, number of defective items, patients per shift, cars in a parking lot, emails received per day.

Continuous variables can take any value within a range, including decimals and fractions. Between any two measurements, there is always another possible value. The precision is limited only by your measuring instrument.

Examples: height, weight, temperature, time, blood pressure, distance.

The key distinction: Can there be values between any two data points? If yes, the variable is continuous. If the values jump from one whole number to the next with nothing in between, it is discrete.

Example 2: Identifying Variable Types

A retail store tracks these metrics. Classify each as discrete or continuous.

Number of complaints is a classic discrete variable. You can only have whole counts — there is no such thing as 2.7 complaints in a day.

Levels of Measurement

The levels of measurement are the most important classification system in statistics because they determine exactly which operations and statistics are valid for your data. There are four levels, each building on the one below it.

Nominal

Nominal data consists of categories with no natural order. You can group items and count how many fall into each category, but you cannot rank them.

Examples: blood type (A, B, AB, O), gender, brand preference, country of birth, eye color.

Allowed operations: Count, mode, frequency tables. That is all.

You can say “more patients have Type O than Type AB,” but you cannot say Type O is “greater than” Type AB — the categories have no inherent ranking.

Ordinal

Ordinal data has categories with a meaningful order, but the intervals between categories are not necessarily equal.

Examples: pain scale (0 to 10), education level (high school, bachelor’s, master’s, doctorate), Likert scale responses (strongly disagree, disagree, neutral, agree, strongly agree), military rank, class rank.

Allowed operations: Everything nominal allows, plus ranking, median, and percentiles.

The critical limitation: you can say a master’s degree is “higher” than a bachelor’s, but you cannot say the difference between a bachelor’s and a master’s is the same as the difference between a master’s and a doctorate. The intervals are unequal or undefined.

Interval

Interval data is ordered with equal intervals between values, but there is no true zero point. A value of zero does not mean “none” of the quantity.

Examples: temperature in °F or °C (0°F does not mean “no temperature”), calendar year (year 0 is not “no time”), IQ score, SAT score.

Allowed operations: Everything ordinal allows, plus addition, subtraction, and mean.

You can say that the difference between 60°F and 80°F is the same as the difference between 30°F and 50°F (both are 20°F). But you cannot say 80°F is “twice as hot” as 40°F, because the zero point is arbitrary.

Ratio

Ratio data has all the properties of interval data plus a true, meaningful zero. Zero means “none” of the quantity — zero weight means no weight, zero dollars means no money, zero seconds means no time elapsed.

Examples: weight, height, income, age, time, distance, number of items, blood pressure.

Allowed operations: Everything interval allows, plus multiplication, division, and ratios. You can say “Person A earns twice as much as Person B.”

Reference Table

Each level inherits all properties of the levels below it. Ratio data is also interval, ordinal, and nominal. Interval data is also ordinal and nominal. This means that any analysis valid for a lower level is also valid for a higher level — you can always find the mode of ratio data, for instance — but you cannot apply higher-level operations to lower-level data.

Example 3: Classifying Survey Questions

A workplace satisfaction survey collects the following data. Classify each by level of measurement.

A common mistake is classifying satisfaction ratings as interval data. While the numbers 1 through 5 appear equally spaced, we have no evidence that the psychological difference between “strongly disagree” and “disagree” is the same as between “agree” and “strongly agree.” This is why most statisticians treat Likert-scale data as ordinal.

Why Data Type Matters

Choosing the wrong analysis for your data type leads to meaningless or misleading results. The following table summarizes which visualizations and statistics are appropriate for each type.

For example, calculating the mean of nominal data like zip codes produces a number, but that number is garbage — there is no “average zip code.” Similarly, creating a histogram of categorical data like blood types makes no sense because the x-axis has no meaningful order or scale.

When in doubt, start by identifying the level of measurement. That tells you exactly which statistics and graphs are valid.

Real-World Application: Nursing — Classifying Patient Data

A nurse reviews a patient intake form with the following fields. Each must be classified by data type and level of measurement to determine how it can be analyzed.

Why this matters in practice:

• You can calculate the mean age and mean weight of patients on a ward because both are ratio-level data. You cannot calculate a “mean gender” or “mean insurance provider.”
• You can rank pain levels — a patient at level 8 is in more pain than one at level 3. But you should not assume the difference between 3 and 4 equals the difference between 7 and 8. The clinical gap between moderate and severe pain may be very different from the gap between mild and moderate.
• Temperature in Fahrenheit is interval, not ratio. You can say that the difference between 98.6°F and 100.6°F is the same as the difference between 100.6°F and 102.6°F (both 2°F). But you cannot say 100°F is “twice as hot” as 50°F — that comparison requires the Kelvin scale, which has a true zero.
• Blood pressure category is ordinal. A hospital can report that 40% of patients have “high” BP, but computing a mean of (low, normal, high, critical) requires assigning numeric values — and the choice of those values is subjective, which is why many researchers use the median or mode for ordinal data.

Practice Problems

Test your understanding with these problems. Click to reveal each answer.

Key Takeaways

• Categorical data describes qualities or labels (blood type, eye color). Quantitative data represents measurable amounts (height, weight, score).
• Discrete variables take countable values (number of patients). Continuous variables can take any value in a range (temperature, weight).
• The four levels of measurement — nominal, ordinal, interval, ratio — form a hierarchy where each level adds a property to the one below it.
• Nominal data can only be counted. Ordinal adds ranking. Interval adds equal spacing. Ratio adds a true zero and meaningful ratios.
• The level of measurement determines which statistics are valid: you cannot compute a mean on nominal data or claim meaningful ratios with interval data.
• Numbers are not always quantitative — zip codes, phone numbers, jersey numbers, and room numbers are categorical despite being digits.
• When in doubt, ask: “Does arithmetic on this variable produce a meaningful result?” If not, the data is categorical.

Return to Statistics for more topics in this section.