Beginner6 min readSource-backed

Types of Data: Scales of Measurement

Learn the four scales of measurement — nominal, ordinal, interval, and ratio — and why they determine which statistical test you can use.

You'll learn

How variable types determine which statistical test is valid.

Use this when

Before running any analysis — understanding your variable types is step one.

Why Data Types Matter

Before running any statistical analysis, you must understand what kind of data you have. The scale of measurement determines which summary statistics are meaningful and which statistical tests are valid. Using the wrong test can lead to completely misleading conclusions.

A common mistake is calculating the "average" of nominal data. The average of blood types A, B, AB, and O is meaningless — even though software will happily compute it if you encode them as numbers.

Nominal (Categorical)

Nominal data represents categories with no natural order. The only mathematical operation that makes sense is counting how many observations fall into each category.

●Examples: blood type (A/B/AB/O), sex (male/female), country of birth, treatment group (placebo/drug A/drug B)
●Valid summaries: frequency tables, mode, percentages
●Invalid summaries: mean, median, standard deviation
●Appropriate tests: Chi-square, Fisher exact test

Ordinal (Ranked)

Ordinal data has a meaningful order, but the intervals between values are not equal. You know that "severe pain" is worse than "moderate pain," but not by how much.

●Examples: Likert scales (1–5), education level (high school < bachelor < master < PhD), disease severity (mild/moderate/severe)
●Valid summaries: median, percentiles, mode
●Invalid summaries: mean (debated), standard deviation
●Appropriate tests: Mann-Whitney U, Wilcoxon, Kruskal-Wallis

In practice, many researchers treat 7-point Likert scales as interval data. This is statistically controversial, but widely accepted in social sciences when scales have many points and distributions are symmetric.

Interval (Equal Spacing, No True Zero)

Interval data has equal gaps between values, but zero does not mean "none." Ratios are therefore meaningless — 20°C is not "twice as hot" as 10°C.

●Examples: temperature in Celsius/Fahrenheit, calendar year, IQ scores, pH values
●Valid summaries: mean, standard deviation, range
●Invalid summaries: ratios ("twice as much")
●Appropriate tests: t-tests, ANOVA, correlation, regression

Ratio (True Zero, All Operations Valid)

Ratio data has all the properties of interval data plus a true zero, which means zero genuinely means "none." All arithmetic operations are valid.

●Examples: weight, height, age, income, blood glucose, time, counts
●Valid summaries: mean, SD, geometric mean, coefficient of variation
●Valid ratios: "Twice the dose," "50% reduction in HbA1c"
●Appropriate tests: t-tests, ANOVA, regression, survival analysis

Quick Reference Table

1.Nominal → Can only count categories → Chi-square
2.Ordinal → Can rank, but gaps unequal → Nonparametric tests
3.Interval → Equal gaps, zero is arbitrary → Parametric tests (with care)
4.Ratio → Equal gaps, true zero → All parametric tests

Practice with your own dataset

Upload a dataset with a categorical column (e.g., sex, group, treatment) and run a Frequency Table to see counts and percentages — the natural summary for nominal data.

1.Go to Analyze and upload any dataset
2.Select "Frequency Table" from the analysis menu
3.Pick a categorical column (sex, group, blood type)
4.Review the count and percentage for each category

Open Analyze Ask Research Assistant